CN107958091A - A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping - Google Patents

A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping Download PDF

Info

Publication number
CN107958091A
CN107958091A CN201711459522.8A CN201711459522A CN107958091A CN 107958091 A CN107958091 A CN 107958091A CN 201711459522 A CN201711459522 A CN 201711459522A CN 107958091 A CN107958091 A CN 107958091A
Authority
CN
China
Prior art keywords
financial
nlp
client
information
entity
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
Application number
CN201711459522.8A
Other languages
Chinese (zh)
Inventor
马天平
侯玥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Beta Technology Co Ltd
Original Assignee
Beijing Beta Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Beta Technology Co Ltd filed Critical Beijing Beta Technology Co Ltd
Priority to CN201711459522.8A priority Critical patent/CN107958091A/en
Publication of CN107958091A publication Critical patent/CN107958091A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/06—Asset management; Financial planning or analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Artificial Intelligence (AREA)
  • Databases & Information Systems (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Animal Behavior & Ethology (AREA)
  • Game Theory and Decision Science (AREA)
  • Human Resources & Organizations (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping, pass through the financial vertical collection of illustrative plates of foundation, it is associated with NLP, that is, natural language processing, so as to establish a set of perfect intelligent finance problem interactive system, with traditional search, question answering system is different by dynamic response, such a method can carry out active rhetorical question, recommend, excavate the problem of user's deep layer is wanted to ask, in flow, the problem of such a method receives user, first pass through NLP technologies, then semanteme is analyzed, retrieved by semanteme into financial vertical collection of illustrative plates, draw a circle to approve the scene domain of question answering, choose most suitable answer and be pushed to client, the problem of client can so being proposed, problem is better understood from reference to knowledge mapping, it can more accurately retrieve answer at the same time.

Description

A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping
Technical field
The present invention relates to artificial intelligence conversational system field, robot language field.
Background technology
At present, financing manager generally serves as the role of shopping guide, client is whole mostly in the purchasing process of financial product Lasting communication is needed to accompany during decision thinking.And it is different from traditional product sale, for financial product, purchase is only It is a beginning, subsequent process needs more link up to be accompanied to play the role of looking after.But using manual service come whole day Wait and communication is provided, ensure that the service of high-quality and high-efficiency is clearly impossible.The rise of AI intelligence so as to the round-the-clock height of client What effect intelligent Service was technically realized becomes possibility.
And in the prior art, the technology that interactive system is relied on is broadly divided into following two：
Based on the technology in question and answer storehouse, i.e., carried out by extracting the keyword in sentence into the magnanimity question and answer put in order Match somebody with somebody.Many intelligent customer services are exactly the cost adopted this method instead of the artificial customer service in part, and this method shortcoming is sea Measure between data without interrelated, if in addition sentence word order in part is replaced, the meaning is entirely different, but matched answer is likely to be one Cause；
Technology based on search engine, i.e., screened the result returned the problem of user using search engine and returned Answer, and on the one hand answer quality that this technology obtains is low, the answer on the other hand obtained is uncontrollable.
As it can be seen that at least there are problems that in the prior art it is following some：
1. the problem of conventional method and corresponding answer are relatively independent storages, onrelevant between each problem, can not be effective Customer problem is speculated by semanteme；
2. conventional method be more lateral to solve the problems, such as frequent user, it is pre-sales the problem of, it is most of personalized for after sale Problem can not be answered accurately；
3. matching of the conventional method based on template, problem richness manually get by the extensive of magnanimity problem, same to ask Topic needs to expect various ways to put questions, it is necessary to take a substantial amount of time and energy.
The content of the invention
For problems of the prior art, this application provides a kind of brand-new solution.
This application discloses a kind of NLP (natural language processing) artificial intelligence approaches based on financial vertical knowledge mapping with Realize the intelligent finance problem interaction between machine and client, it specifically includes following steps：
S1, enquirement, client propose the problem of finance is related；
S2, NLP natural language processing, the problem of being proposed by NLP technologies to client, are handled；
S3, semantic analysis and understanding, according to suitable phrase, the key for representing entity/relation of information selection after processing Word is in case later retrieval uses；
S4, information retrieval, according to the semantic analysis with understanding that obtained information is retrieved in financial vertical collection of illustrative plates；
S5, result output, answer output is generated to answer client questions according to retrieval result, and carry out problem guiding and Excavate.
Wherein, the financial vertical collection of illustrative plates is made by data base manipulation machine learning techniques by financial field knowledge Relation is combed and is preserved and obtained between the financial field knowledge.
Further, the making of the financial vertical collection of illustrative plates includes：
1) financial field relational learning data is prepared；
2) the financial field relational learning data is learnt using semi-supervised mode by machine learning techniques；
3) relation between the financial field knowledge entity learnt and financial field knowledge entity is combed and protected Deposit and obtain chart database in the database；
4) semi-supervised maintenance is carried out to the relation between the financial field knowledge entity, will represents the synonymous of the same relation Word is polymerize in the range of the entity that the relation is related to；
5) selective extraction uses when describing to common entity increase picture, and being exported as search terms for subsequent result, The variation of interactive answer form can so be improved.
Wherein, when each entitative concept deposit, other existing entities can be associated, each inter-entity is ultimately formed and exists The financial vertical collection of illustrative plates of correlation.
Further, the NLP natural language processings sequentially comprise the following steps：
S2.1 data processings, the data processing include complicated and simple conversion, Chinese word segmentation, part-of-speech tagging, data cleansing, syntax Parsing, Entity recognition and/or voice turn word process；
S2.2 disaggregated models, the disaggregated model is in terms of problem types, user behavior, emotion recognition three to the number Classify according to processed problem；
S2.3 information extractions, described information extraction is pressed is extracted, based on a language piece based on part-of-speech tagging extraction, based on semantic analysis The step of analysis extraction, carries out successively；
Wherein described word-based property is marked extraction and is referred to as extracted based on PoS (Part-of-Speech), i.e., basis marks out Noun, verb, adjective or other parts of speech, make choice extraction；
It is described that the carrying into row information of the mode based on industry keyword or list of entities is then selected based on semantic analysis extraction Calling is taken, the method for calling is exactly the statistical methods such as common card side well known in the industry, information gain, mutual information technology.
The Discourse anlysis then carries out extraction calling by the way of dependency tree to information, wherein the dependency tree represents sentence In dependence between each word.
S2.4 information completions, detect the missing of sentence structure in client's proposition problem, and by lack part completion；
S2.5 message queues：The key message of extraction is put into queue, and again by the key message after more than 5 wheel dialogues Queue is removed, while resolution paraphrase also is carried out to the demonstrative pronoun in sentence.
Wherein dialogue and queue refer specifically to, and the key message of nearest a word that user is said is put into this information team In row, the relatively early key message into enqueue (such as before 5 wheels) will be moved out of queue.The queuing message supplies the language of information retrieval Adopted scope is controlled, and more wheel problems are replied and provide foundation.
Further, described information retrieval sequentially comprises the following steps：
S4.1 Question Classifications, according to semantic analysis with understanding scene of the obtained information to Question Classification delineation question answering Scope, the type of described problem include：Special screne class problem, throw problem, QA (question and answer class) problem after preceding or throwing.
S4.2 problems are retrieved, in the financial vertical collection of illustrative plates in the scene domain drawn a circle to approve after problem types has been determined The middle information obtained according to semantic analysis with understanding is retrieved to obtain required entity and/or relation, is total to afterwards according to word Show similarity, sentence pattern matching strategy and/or generate answer in case being sent to client with the mode that word order is considered.
Word co-occurrence similarity is a kind of model based on statistics, in one section of language expression, is often had some There is (i.e. co-occurrence) in same sentence or same paragraph jointly in a word, is so considered as these words in the sense It is related.Similarity measure is carried out using such a mode, is the entity that will be come out by semantic retrieval, passes through such a side Formula is calculated, and improves the accuracy of follow-up output result.
The effect of sentence pattern matching strategy is to be extracted the related entities needed for replying by information retrieval step, but also not It is a complete sentence.By sentence pattern matching strategy, can by it is known the problem of clause, it is corresponding to choose the clause answered, The generation that " filling of blank groove position " of re-multiplexing " information completion " part is answered.
It is one of co-occurrence similarity supplement to be considered with word order, after entity is chosen, really have chosen one group it is similar Entity, by the group object and the matching degree of problem retrieved, so that it is determined that the priority ranking of the group object, preferentially Level is high to be appeared in answer, and low answers to be alternative.
Further, result output includes the answer of generation, guides and the problem of excavating is exported to client, its Described in the problem of guiding and excavating for according to entity associated with the answer in the financial vertical collection of illustrative plates into advancing one Step guiding and excavate, forming the problem of new exports and answered or confirmed for client to client, can so realize truer Effectively talk with interaction mode.
In addition, utilize the NLP artificial intelligence approaches progress based on financial vertical knowledge mapping disclosed herein as well is a kind of The system of intelligent finance problem interaction, specifically includes input module, NLP natural language processings module, semantic analysis and understands mould Block, information searching module and result output module；
Wherein,
Described problem is simultaneously sent to the NLP by the problem of finance correlation that the input module is used to receive client's proposition Natural language processing module is handled；
The problem of NLP natural language processings module proposes client by NLP technologies is handled；
The semantic analysis and Understanding Module according to the suitable phrase for representing entity/relation of information selection after processing, Keyword is in case later retrieval uses；
Described information retrieves module according to the semantic analysis with understanding obtained information in the financial vertical collection of illustrative plates Retrieved；
The result output module generates answer output to answer client questions according to retrieval result, and carries out drawing for problem Lead and excavate.
By the financial vertical collection of illustrative plates of foundation, it is associated with NLP (natural language processing), so as to establish a set of perfect Intelligent finance problem interactive system.Different by dynamic response from traditional search, question answering system, such a method can carry out actively anti- Ask, recommend, excavate the problem of user's deep layer is wanted to ask.In flow, the problem of such a method receives user, NLP technologies are first passed through, so Post analysis go out semanteme, are retrieved by semanteme into financial vertical collection of illustrative plates, draw a circle to approve the scene domain of question answering, choose most suitable The answer of conjunction is pushed to client.
The problem of one side can propose client is so done, problem is better understood from reference to knowledge mapping, on the other hand Answer can more accurately be retrieved.There are many knowledge points in whole finance activities, comb the relation of involved knowledge point, Appropriate is stored by chart database, in case NLP systems carry out follow-up reading retrieval.
The application compared with the prior art, the advantage is that：
1. by the financial knowledge mapping of foundation, can with entity/relation data in constantly improve and abundant data storehouse, When each entitative concept deposit, other existing entities can be intelligently associated, ultimately forming one, there are complex relationship Financial vertical collection of illustrative plates, so as to tackle the personalized question of client's complexity, while deposit (such as the picture of different types of data Information) and distinctive information retrieval and answer generation output mechanism can improve the variation of interactive answer form.
2. active rhetorical question can be carried out, recommend, excavate the problem of user's deep layer is wanted to ask, friendship is talked with realization more authentic and validly Mutual state.
3. understanding customer issue that can be more preferably more efficient, and most suitable answer is accurately retrieved, greatly improved pair Talk about communication efficiency.
Brief description of the drawings
Fig. 1 is the exemplary explanatory drawin of financial knowledge entity in the financial vertical collection of illustrative plates that diagram data place is deposited；
Fig. 2 is the overall procedure schematic diagram of Intelligent dialogue；
Fig. 3 is to realize the intelligence between machine and client based on the NLP artificial intelligence approaches of financial vertical knowledge mapping The detail flowchart of monetary affair interaction.
Embodiment
For abundant disclosed purpose, the present invention is described in further details below with reference to embodiment.It should be appreciated that with The lower specific embodiment is only used for explaining the present invention, is not intended to limit the scope of the present invention.
The application specifically disclose it is a kind of based on the NLP artificial intelligence approaches of financial vertical knowledge mapping with realize machine and Intelligent finance problem interaction between client, referring to Fig. 2, it specifically includes following steps：
S1, enquirement, client propose the problem of finance is related；
S2, NLP natural language processing, the problem of being proposed by NLP technologies to client, are handled；
S3, semantic analysis and understanding, according to suitable phrase, the key for representing entity/relation of information selection after processing Word is in case later retrieval uses；
S4, information retrieval, according to the semantic analysis with understanding that obtained information is retrieved in financial vertical collection of illustrative plates；
S5, result output, answer output is generated to answer client questions according to retrieval result, and carry out problem guiding and Excavate.
Wherein, the financial vertical collection of illustrative plates is made by data base manipulation machine learning techniques by financial field knowledge Relation is combed and is preserved and obtained between the financial field knowledge.
The example of financial knowledge entity, the making bag of the financial vertical collection of illustrative plates in financial vertical collection of illustrative plates as shown in Figure 1 Include：
1) financial field relational learning data is prepared, the learning materials are bank's relevant knowledges in the example depicted in fig. 1 Data；
2) the financial field relational learning data is learnt using semi-supervised mode by machine learning techniques, There is the information of Chinese four big rows and its dependency relation in the example, in learning materials；
3) relation between the financial field knowledge entity learnt and financial field knowledge entity is combed and protected Deposit and obtain chart database in the database, in this example, study " Chinese four great Hang You Bank of Chinas, the Industrial and Commercial Bank of China, in Agricultural bank of state, China Construction Bank ", by this study to 5 entities and 4 groups of relations, i.e., Chinese four big rows-have-in Bank of state, the Chinese four big rows-have-Industrial and Commercial Bank of China, the Chinese four big rows-have-Agricultural Bank of China, Chinese four big rows-have-in Construction Bank of state；
4) semi-supervised maintenance is carried out to the relation between the financial field knowledge entity, will represents the synonymous of the same relation Word is polymerize in the range of the entity that the relation is related to, such as, financial knowledge is received again, and Chinese four big rows include China Bank, the Industrial and Commercial Bank of China, the Agricultural Bank of China, China Construction Bank.Already present 5 inter-entity new relation can occur again "comprising", and it is consistent with "comprising" in semantically existing two relations " having ", a relation can be merged into, and the entity is remembered Relation similar in record, improves the accuracy of machine learning；
5) to common entity increase picture description, and used as search terms for subsequently selected extraction, the deposit of picture In order to improve the diversification of forms of interaction answer, relevant picture is imported to common entity, selective extraction when for replying.
Wherein, when each entitative concept deposit, other existing entities can be associated, each inter-entity is ultimately formed and exists The financial vertical collection of illustrative plates of correlation.
As shown in figure 3, Fig. 3 show the NLP artificial intelligence approaches based on financial vertical knowledge mapping with realize machine and The detailed process of intelligent finance problem interaction between client,
Wherein NLP natural language processings sequentially comprise the following steps described in step S2：
S2.1 data processings, the data processing include complicated and simple conversion, Chinese word segmentation, part-of-speech tagging, data cleansing, syntax Parsing, Entity recognition and/or voice turn word process；
Wherein Chinese word segmentation use the mathematic(al) representation of method for：
(T)=(W1)P(W2|W1)P(W3|W2)…P(Wn|Wn-1)
P represents the probability occurred, wherein W1Represent first character region recognition as a result, so P (W1) represent first character The probability of appearance.P(W2|W1) if first character appearance is represented, then there is the probability of the second word, it can be seen that W1W2Whether be A common word.So operation will generate a sentence for having divided word until sentence end.Such as：China | four is big | silver OK | which has |
S2.2 disaggregated models, the disaggregated model is in terms of problem types, user behavior, emotion recognition three to the number Classify according to processed problem；
Problem types be mainly to discriminate between the problem of user be 5w classes problem (where, what, when, why, how), it is conventional Problem also matter of right and wrong class is linked up.
User behavior refers to the enquirement that user carries out under which page, what environment.
Emotion recognition can be involved according to user this conversation procedure, extraction emotion word.
S2.3 information extractions, described information extraction is pressed is extracted, based on a language piece based on part-of-speech tagging extraction, based on semantic analysis The step of analysis extraction, carries out successively；
Wherein described word-based property is marked extraction and is referred to as extracted based on PoS (Part-of-Speech), i.e., basis marks out Noun, verb, adjective or other parts of speech, make choice extraction；
It is described that the carrying into row information of the mode based on industry keyword or list of entities is then selected based on semantic analysis extraction Calling is taken, the method for calling is exactly the statistical methods such as common card side well known in the industry, information gain, mutual information technology.
The Discourse anlysis then carries out extraction calling by the way of dependency tree to information, wherein the dependency tree represents sentence In dependence between each word.
More specifically, the mathematic(al) representation of information extraction is wherein carried out by card side：
Feature extraction algorithm is divided into two major class of feature selecting and feature extraction.Wherein Chi-square Test just belongs to feature selecting calculation Preferably algorithm in method.
T and c is two stochastic variables respectively, and χ 2 represents chi-square value, and to the correlation of inspection data, t represents a word, C represents a classification.For example t can represent flower, c can represent plant.
When we have looked for N articles, the classification respectively included t, not included, then c belonged to, is not belonging to Classification.So as to generate below table, the letter in above-mentioned formula is corresponded to respectively.
Feature selecting C1. belong to " plant " C2. it is not belonging to " plant " Amount to
T1. " flower " is included A B A+B
T2. " flower " is not included C D C+D
Sum A+C B+D N
The mathematic(al) representation of information extraction is wherein carried out by information gain：
In the case where being characterized as that Y is fixed, the conditional entropy of X is H (X | Y), and P (x | y) is the probability that occurs.Popular point is explained It is exactly, when the uncertainty of information content X after knowing this information content of Y is not compared to information content Y is known, the uncertainty of X Reduce how many.Two variable Xs of information gain, the status of Y is different, is that Y is regarded as to reduce a kind of probabilistic means of X. And two variable status of mutual information are identical.
The mathematic(al) representation of information extraction is wherein carried out by mutual information：
G (D, A)=H (D)-H (D | A)
H (D) represents self-information amount, that is, the information sent, and H (D | A) conditional information content is represented, g (D, A) represents mutual information Amount.So mutual information=self-information amount-conditional information content.Information extraction is carried out by mutual information, word is calculated using mutual information The internal bond strength of string so that the information of extraction is more complete, rather than the state of keyword.Such as natural language processing, It finds that these words are a fixed academic vocabulary by mutual information, so three words can't be split as to understand.
S2.4 information completions, detect the missing (also referred to as groove position blank detection) of sentence structure in client's proposition problem, and By lack part completion (also referred to as blank groove position filling)；
S2.5 message queues：The key message of extraction is put into queue, and again by the key message after more than 5 wheel dialogues Queue is removed, while resolution paraphrase also is carried out to the demonstrative pronoun in sentence.
Wherein dialogue and queue refer specifically to, and the key message of nearest a word that user is said is put into this information team In row, the relatively early key message into enqueue (such as before 5 wheels) will be moved out of queue.The queuing message supplies the language of information retrieval Adopted scope is controlled, and more wheel problems are replied and provide foundation.
Wherein step S4 described informations retrieval sequentially comprises the following steps：
S4.1 Question Classifications, according to semantic analysis with understanding scene of the obtained information to Question Classification delineation question answering Scope, the type of described problem include：Special screne class problem, throw problem, QA (question and answer class) problem after preceding or throwing.
S4.2 problems are retrieved, in the financial vertical collection of illustrative plates in the scene domain drawn a circle to approve after problem types has been determined The middle information obtained according to semantic analysis with understanding is retrieved to obtain required entity and/or relation, is total to afterwards according to word Show similarity, sentence pattern matching strategy and/or generate answer in case being sent to client with the mode that word order is considered.
Co-occurrence similarity measure mathematic(al) representation：
Similarity (s1, s2)=α SDMG (s1, s2)+β SDMG (s1, s2)+γ SDMG (s1, s2)
S1, s2 are two words, and α, β, γ are the co-occurrence correlations in three space vectors respectively.
Result output includes the answer of generation, guides and the problem of excavating is exported to client wherein described in step S4, Wherein described guiding and the problem of excavate for according to entity associated with the answer in the financial vertical collection of illustrative plates into traveling One step is guided and excavated, and is formed the problem of new and is exported and answered or confirmed for client to client, can so realize trueer It is real effectively to talk with interaction mode., such as：
End user asks：Which Chinese Big Four (Barclay Bank, Lloyd's Bank, Midland Bank, National Westinster Bank) has
Dialogue is answered：Chinese Big Four (Barclay Bank, Lloyd's Bank, Midland Bank, National Westinster Bank) Dou You Bank of Chinas, the Industrial and Commercial Bank of China, the Agricultural Bank of China, China Reconstructs Bank, did you knowWherein the Industrial and Commercial Bank of China set up in 1984.
Above-described embodiment specifically can carry out intelligence by using the NLP artificial intelligence approaches based on financial vertical knowledge mapping Can the system of monetary affair interaction realize that the system comprises input module, NLP natural language processings module, semantic analysis With Understanding Module, information searching module and result output module；
Wherein,
Described problem is simultaneously sent to the NLP by the problem of finance correlation that the input module is used to receive client's proposition Natural language processing module is handled；
The problem of NLP natural language processings module proposes client by NLP technologies is handled；
The semantic analysis and Understanding Module according to the suitable phrase for representing entity/relation of information selection after processing, Keyword is in case later retrieval uses；
Described information retrieves module according to the semantic analysis with understanding obtained information in the financial vertical collection of illustrative plates Retrieved；
The result output module generates answer output to answer client questions according to retrieval result, and carries out drawing for problem Lead and excavate.
The above implements the embodiment that thin example only expresses this patent, it is described it is not intended that to this patent scope Limitation.It should be pointed out that for those of ordinary skill in the art, on the premise of this patent design is not departed from, Various modifications and improvements can be made, these belong to protection scope of the present invention.Therefore, the protection domain of this patent should be with Subject to appended claims.Embodiment described above only expresses embodiments of the present invention, its description is more specific and detailed, But therefore it can not be interpreted as the limitation to the scope of the claims of the present invention.It should be pointed out that the ordinary skill for this area For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the present invention Protection domain.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.

Claims (6)

1. a kind of NLP artificial intelligence approaches based on financial vertical knowledge mapping, it is characterised in that it specifically includes following step Suddenly：
S1, enquirement, client propose the problem of finance is related；
S2, NLP natural language processing, the problem of being proposed by NLP technologies to client, are handled；
S3, semantic analysis and understanding, the phrase of entity/relation, keyword are represented in case follow-up according to the information selection after processing Retrieval uses；
S4, information retrieval, according to the semantic analysis with understanding that obtained information is retrieved in financial vertical collection of illustrative plates；
S5, result output, answer output is generated to answer client questions according to retrieval result, and carries out the guiding and digging of problem Pick；
Wherein, the financial vertical collection of illustrative plates is made by data base manipulation machine learning techniques by financial field knowledge and institute Relation is combed and is preserved and obtained between stating financial field knowledge.
2. the NLP artificial intelligence approaches based on financial vertical knowledge mapping as described in claim 1, it is characterised in that institute Stating the making of financial vertical collection of illustrative plates includes：
1) financial field relational learning data is prepared；
2) the financial field relational learning data is learnt using semi-supervised mode by machine learning techniques；
3) relation between the financial field knowledge entity learnt and financial field knowledge entity is combed and is stored in Chart database is obtained in database；
4) semi-supervised maintenance is carried out to the relation between the financial field knowledge entity, the synonym for representing the same relation is existed It is polymerize in the range of the entity that the relation is related to；
5) selective extraction uses when describing to common entity increase picture, and being exported as search terms for subsequent result, so The variation of interactive answer form can be improved；
Wherein, when each entitative concept deposit, other existing entities can be associated, each inter-entity is ultimately formed and exists mutually The financial vertical collection of illustrative plates of relation.
3. the NLP artificial intelligence approaches based on financial vertical knowledge mapping as described in claim 1, it is characterised in that institute NLP natural language processings are stated sequentially to comprise the following steps：
S2.1 data processings, the data processing include complicated and simple conversion, Chinese word segmentation, part-of-speech tagging, data cleansing, syntax solution Analysis, Entity recognition and/or voice turn word process；
S2.2 disaggregated models, the disaggregated model in terms of problem types, user behavior, the emotion recognition three to the data at The problem of managing is classified；
S2.3 information extractions, described information extraction is pressed is extracted, based on Discourse anlysis based on part-of-speech tagging extraction, based on semantic analysis The step of extraction, carries out successively；
S2.4 information completions, detect the missing of sentence structure in client's proposition problem, and by lack part completion；
S2.5 message queues：The key message of extraction is put into queue, and again removes the key message after more than 5 wheel dialogues Queue, while resolution paraphrase also is carried out to the demonstrative pronoun in sentence.
4. the NLP artificial intelligence approaches based on financial vertical knowledge mapping as described in claim 1, it is characterised in that institute Information retrieval is stated sequentially to comprise the following steps：
S4.1 Question Classifications, according to semantic analysis with understanding scene model of the obtained information to Question Classification delineation question answering Enclose, the type of described problem includes：Special screne class problem, throw problem, QA (question and answer class) problem after preceding or throwing；
S4.2 problems retrieve, after having determined problem types in the scene domain drawn a circle to approve in the financial vertical collection of illustrative plates root According to semantic analysis with understanding that obtained information is retrieved to obtain required entity and/or relation, afterwards according to word co-occurrence phase Answer is generated in case being sent to client like degree, sentence pattern matching strategy and/or with the mode that word order is considered.
5. the NLP artificial intelligence approaches based on financial vertical knowledge mapping as described in claim 1, it is characterised in that institute Stating result output includes the answer of generation, guides and the problem of excavating is exported to client, wherein the guiding and excavating The problem of according to the entity associated with the answer in the financial vertical collection of illustrative plates further guide and excavate, formed new The problem of export and answered or confirmed for client to client.
6. in a kind of 1-5 using claim it is any it is described based on the NLP artificial intelligence approaches of financial vertical knowledge mapping into Row intelligent finance problem interaction system, it is characterised in that the system comprises input module, NLP natural language processings module, Semantic analysis and Understanding Module, information searching module and result output module；
Wherein,
Described problem is simultaneously sent to the NLP natures by the problem of finance correlation that the input module is used to receive client's proposition Language processing module is handled；
The problem of NLP natural language processings module proposes client by NLP technologies is handled；
The semantic analysis is with Understanding Module according to suitable phrase, the key for representing entity/relation of information selection after processing Word is in case later retrieval uses；
Described information retrieves module according to the semantic analysis with understanding that obtained information carries out in the financial vertical collection of illustrative plates Retrieval；
The result output module generates answer output to answer client questions according to retrieval result, and carry out problem guiding and Excavate.
CN201711459522.8A 2017-12-28 2017-12-28 A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping Pending CN107958091A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711459522.8A CN107958091A (en) 2017-12-28 2017-12-28 A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711459522.8A CN107958091A (en) 2017-12-28 2017-12-28 A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping

Publications (1)

Publication Number Publication Date
CN107958091A true CN107958091A (en) 2018-04-24

Family

ID=61957098

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711459522.8A Pending CN107958091A (en) 2017-12-28 2017-12-28 A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping

Country Status (1)

Country Link
CN (1) CN107958091A (en)

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108733654A (en) * 2018-05-21 2018-11-02 宁波薄言信息技术有限公司 A kind of information processing method
CN108920603A (en) * 2018-06-28 2018-11-30 厦门快商通信息技术有限公司 A kind of customer service bootstrap technique based on customer service machine mould
CN108932340A (en) * 2018-07-13 2018-12-04 华融融通（北京）科技有限公司 The construction method of financial knowledge mapping under a kind of non-performing asset operation field
CN109344174A (en) * 2018-09-13 2019-02-15 深圳易投云智能科技有限公司 Financial analysis method and system
CN109559144A (en) * 2018-09-20 2019-04-02 欧阳江 A kind of personalization securities industry customer service system and method
CN109658271A (en) * 2018-12-19 2019-04-19 前海企保科技（深圳）有限公司 A kind of intelligent customer service system and method based on the professional scene of insurance
CN109829041A (en) * 2018-12-25 2019-05-31 出门问问信息科技有限公司 Question processing method and device, computer equipment and computer readable storage medium
CN109918489A (en) * 2019-02-28 2019-06-21 上海乐言信息科技有限公司 A kind of knowledge question answering method and system of more strategy fusions
CN110019687A (en) * 2019-04-11 2019-07-16 宁波深擎信息科技有限公司 A kind of more intention assessment systems, method, equipment and the medium of knowledge based map
CN110019742A (en) * 2018-06-19 2019-07-16 北京京东尚科信息技术有限公司 Method and apparatus for handling information
CN110443382A (en) * 2019-07-09 2019-11-12 青岛聚好联科技有限公司 A kind of user interaction approach based on intelligent customer service platform
CN110543546A (en) * 2019-07-18 2019-12-06 华瑞新智科技（北京）有限公司 Method and device for inquiring hidden danger problems in industrial equipment
CN110597952A (en) * 2019-08-20 2019-12-20 深圳壹账通智能科技有限公司 Information processing method, server, and computer storage medium
CN110727773A (en) * 2019-10-11 2020-01-24 沈阳民航东北凯亚有限公司 Information providing method and device
CN110807091A (en) * 2019-03-01 2020-02-18 王涵 Hotel intelligent question-answer recommendation and decision support analysis method and system
CN110990526A (en) * 2019-11-21 2020-04-10 腾讯科技（深圳）有限公司 Query statement display method and related equipment
CN111091454A (en) * 2019-11-05 2020-05-01 新华智云科技有限公司 Financial public opinion recommendation method based on knowledge graph
WO2020114269A1 (en) * 2018-12-05 2020-06-11 中兴通讯股份有限公司 Robo-advisor implementation method and system
CN111309888A (en) * 2020-02-25 2020-06-19 百度在线网络技术（北京）有限公司 Man-machine conversation method, device, electronic equipment and storage medium
CN111739541A (en) * 2019-03-19 2020-10-02 上海云思智慧信息技术有限公司 Conference assistance method and system based on voice, storage medium and terminal
CN111753021A (en) * 2020-06-17 2020-10-09 第四范式（北京）技术有限公司 Method, device and equipment for constructing knowledge graph and readable storage medium
WO2020221142A1 (en) * 2019-04-28 2020-11-05 华为技术有限公司 Picture book-based question and answer interaction method and electronic device
CN112015919A (en) * 2020-09-15 2020-12-01 重庆广播电视大学重庆工商职业学院 Dialogue management method based on learning auxiliary knowledge graph
CN112328808A (en) * 2020-11-03 2021-02-05 四川长虹电器股份有限公司 Knowledge graph-based question and answer method and device, electronic equipment and storage medium
CN113312489A (en) * 2021-04-13 2021-08-27 武汉烽火众智数字技术有限责任公司 Panorama retrieval system and method based on NLP and graph database
CN113326367A (en) * 2021-06-30 2021-08-31 四川启睿克科技有限公司 Task type dialogue method and system based on end-to-end text generation
CN117151122A (en) * 2023-10-30 2023-12-01 湖南三湘银行股份有限公司 Bank customer service session question-answering processing method and system based on natural language processing

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104050256A (en) * 2014-06-13 2014-09-17 西安蒜泥电子科技有限责任公司 Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method
CN105868313A (en) * 2016-03-25 2016-08-17 浙江大学 Mapping knowledge domain questioning and answering system and method based on template matching technique
CN106776711A (en) * 2016-11-14 2017-05-31 浙江大学 A kind of Chinese medical knowledge mapping construction method based on deep learning
CN106919655A (en) * 2017-01-24 2017-07-04 网易（杭州）网络有限公司 A kind of answer provides method and apparatus
CN107341215A (en) * 2017-06-07 2017-11-10 北京航空航天大学 A kind of vertical knowledge mapping classification ensemble querying method of multi-source based on Distributed Computing Platform

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104050256A (en) * 2014-06-13 2014-09-17 西安蒜泥电子科技有限责任公司 Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method
CN105868313A (en) * 2016-03-25 2016-08-17 浙江大学 Mapping knowledge domain questioning and answering system and method based on template matching technique
CN106776711A (en) * 2016-11-14 2017-05-31 浙江大学 A kind of Chinese medical knowledge mapping construction method based on deep learning
CN106919655A (en) * 2017-01-24 2017-07-04 网易（杭州）网络有限公司 A kind of answer provides method and apparatus
CN107341215A (en) * 2017-06-07 2017-11-10 北京航空航天大学 A kind of vertical knowledge mapping classification ensemble querying method of multi-source based on Distributed Computing Platform

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
公安部第三研究所: "多摄像机协同关注目标检测跟踪技术", 《多摄像机协同关注目标检测跟踪技术》 *
陆刚等: "机电工程网络信息资源检索与利用", 《机电工程网络信息资源检索与利用》 *

Cited By (38)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108733654A (en) * 2018-05-21 2018-11-02 宁波薄言信息技术有限公司 A kind of information processing method
CN110019742B (en) * 2018-06-19 2024-05-17 北京京东尚科信息技术有限公司 Method and device for processing information
CN110019742A (en) * 2018-06-19 2019-07-16 北京京东尚科信息技术有限公司 Method and apparatus for handling information
CN108920603A (en) * 2018-06-28 2018-11-30 厦门快商通信息技术有限公司 A kind of customer service bootstrap technique based on customer service machine mould
CN108920603B (en) * 2018-06-28 2021-12-21 厦门快商通信息技术有限公司 Customer service guiding method based on customer service machine model
CN108932340A (en) * 2018-07-13 2018-12-04 华融融通（北京）科技有限公司 The construction method of financial knowledge mapping under a kind of non-performing asset operation field
CN109344174A (en) * 2018-09-13 2019-02-15 深圳易投云智能科技有限公司 Financial analysis method and system
CN109559144A (en) * 2018-09-20 2019-04-02 欧阳江 A kind of personalization securities industry customer service system and method
WO2020114269A1 (en) * 2018-12-05 2020-06-11 中兴通讯股份有限公司 Robo-advisor implementation method and system
CN111353013A (en) * 2018-12-05 2020-06-30 中兴通讯股份有限公司 Method and system for realizing intelligent delivery and reception
CN109658271A (en) * 2018-12-19 2019-04-19 前海企保科技（深圳）有限公司 A kind of intelligent customer service system and method based on the professional scene of insurance
CN109829041A (en) * 2018-12-25 2019-05-31 出门问问信息科技有限公司 Question processing method and device, computer equipment and computer readable storage medium
CN109918489A (en) * 2019-02-28 2019-06-21 上海乐言信息科技有限公司 A kind of knowledge question answering method and system of more strategy fusions
CN110807091A (en) * 2019-03-01 2020-02-18 王涵 Hotel intelligent question-answer recommendation and decision support analysis method and system
CN110807091B (en) * 2019-03-01 2023-08-18 王涵 Hotel intelligent question-answer recommendation and decision support analysis method and system
CN111739541A (en) * 2019-03-19 2020-10-02 上海云思智慧信息技术有限公司 Conference assistance method and system based on voice, storage medium and terminal
CN110019687B (en) * 2019-04-11 2021-03-23 宁波深擎信息科技有限公司 Multi-intention recognition system, method, equipment and medium based on knowledge graph
CN110019687A (en) * 2019-04-11 2019-07-16 宁波深擎信息科技有限公司 A kind of more intention assessment systems, method, equipment and the medium of knowledge based map
WO2020221142A1 (en) * 2019-04-28 2020-11-05 华为技术有限公司 Picture book-based question and answer interaction method and electronic device
CN110443382A (en) * 2019-07-09 2019-11-12 青岛聚好联科技有限公司 A kind of user interaction approach based on intelligent customer service platform
CN110543546A (en) * 2019-07-18 2019-12-06 华瑞新智科技（北京）有限公司 Method and device for inquiring hidden danger problems in industrial equipment
CN110543546B (en) * 2019-07-18 2022-03-22 华瑞新智科技（北京）有限公司 Method and device for inquiring hidden danger problems in industrial equipment
CN110597952A (en) * 2019-08-20 2019-12-20 深圳壹账通智能科技有限公司 Information processing method, server, and computer storage medium
CN110727773A (en) * 2019-10-11 2020-01-24 沈阳民航东北凯亚有限公司 Information providing method and device
CN110727773B (en) * 2019-10-11 2022-02-01 沈阳民航东北凯亚有限公司 Information providing method and device
CN111091454A (en) * 2019-11-05 2020-05-01 新华智云科技有限公司 Financial public opinion recommendation method based on knowledge graph
CN110990526B (en) * 2019-11-21 2024-01-30 腾讯科技（深圳）有限公司 Query statement display method and related equipment
CN110990526A (en) * 2019-11-21 2020-04-10 腾讯科技（深圳）有限公司 Query statement display method and related equipment
CN111309888A (en) * 2020-02-25 2020-06-19 百度在线网络技术（北京）有限公司 Man-machine conversation method, device, electronic equipment and storage medium
CN111309888B (en) * 2020-02-25 2023-10-24 百度在线网络技术（北京）有限公司 Man-machine conversation method and device, electronic equipment and storage medium
CN111753021A (en) * 2020-06-17 2020-10-09 第四范式（北京）技术有限公司 Method, device and equipment for constructing knowledge graph and readable storage medium
CN112015919A (en) * 2020-09-15 2020-12-01 重庆广播电视大学重庆工商职业学院 Dialogue management method based on learning auxiliary knowledge graph
CN112328808A (en) * 2020-11-03 2021-02-05 四川长虹电器股份有限公司 Knowledge graph-based question and answer method and device, electronic equipment and storage medium
CN113312489A (en) * 2021-04-13 2021-08-27 武汉烽火众智数字技术有限责任公司 Panorama retrieval system and method based on NLP and graph database
CN113326367A (en) * 2021-06-30 2021-08-31 四川启睿克科技有限公司 Task type dialogue method and system based on end-to-end text generation
CN113326367B (en) * 2021-06-30 2023-06-16 四川启睿克科技有限公司 Task type dialogue method and system based on end-to-end text generation
CN117151122A (en) * 2023-10-30 2023-12-01 湖南三湘银行股份有限公司 Bank customer service session question-answering processing method and system based on natural language processing
CN117151122B (en) * 2023-10-30 2024-03-22 湖南三湘银行股份有限公司 Bank customer service session question-answering processing method and system based on natural language processing

Similar Documents

Publication Publication Date Title
CN107958091A (en) A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping
CN109493166B (en) Construction method for task type dialogue system aiming at e-commerce shopping guide scene
CN110633409B (en) Automobile news event extraction method integrating rules and deep learning
CN110674639B (en) Natural language understanding method based on pre-training model
CN108874782B (en) A kind of more wheel dialogue management methods of level attention LSTM and knowledge mapping
CN104111933B (en) Obtain business object label, set up the method and device of training pattern
CN110472017A (en) A kind of analysis of words art and topic point identify matched method and system
CN108829662A (en) A kind of conversation activity recognition methods and system based on condition random field structuring attention network
CN109271537B (en) Text-to-image generation method and system based on distillation learning
CN107423442A (en) Method and system, storage medium and computer equipment are recommended in application based on user's portrait behavioural analysis
CN107967261A (en) Interactive question semanteme understanding method in intelligent customer service
CN111325029B (en) Text similarity calculation method based on deep learning integrated model
CN105183717B (en) A kind of OSN user feeling analysis methods based on random forest and customer relationship
CN108549658A (en) A kind of deep learning video answering method and system based on the upper attention mechanism of syntactic analysis tree
CN105843897A (en) Vertical domain-oriented intelligent question and answer system
CN104750674B (en) A kind of man-machine conversation's satisfaction degree estimation method and system
CN113505586A (en) Seat-assisted question-answering method and system integrating semantic classification and knowledge graph
CN111753207B (en) Collaborative filtering method for neural map based on comments
CN109087205A (en) Prediction technique and device, the computer equipment and readable storage medium storing program for executing of public opinion index
CN103678318B (en) Multi-word unit extraction method and equipment and artificial neural network training method and equipment
CN105868183A (en) Method and device for predicting staff demission
CN107239489A (en) The prediction of network public-opinion and emulation mode in accident based on SOAR models
CN110188200A (en) A kind of depth microblog emotional analysis method using social context feature
CN112287090A (en) Financial question asking back method and system based on knowledge graph
CN112287106A (en) Online comment emotion classification method based on dual-channel hybrid neural network

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20180424