CN106776832B - Processing method, apparatus and system for question and answer interactive log - Google Patents
Processing method, apparatus and system for question and answer interactive log Download PDFInfo
- Publication number
- CN106776832B CN106776832B CN201611061635.8A CN201611061635A CN106776832B CN 106776832 B CN106776832 B CN 106776832B CN 201611061635 A CN201611061635 A CN 201611061635A CN 106776832 B CN106776832 B CN 106776832B
- Authority
- CN
- China
- Prior art keywords
- answer
- question
- interactive log
- banquet
- attending
- 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.)
- Active
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
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/242—Dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- 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
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Electrically Operated Instructional Devices (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides a kind of processing methods for question and answer interactive log, the question and answer interactive log includes user's question sentence and answer of attending a banquet accordingly, the processing method includes: to execute words art analysis to the question and answer interactive log, prohibits language and/or negative emotion to determine that the answer of attending a banquet whether there is;Business diagnosis is executed to the question and answer interactive log using CRM database, to judge whether user's question sentence consistent with the CRM data of extraction with the business matching and the answer of attending a banquet that user handled;It executes question and answer to the question and answer interactive log using question and answer knowledge base to analyze, to determine whether the answer of attending a banquet is correct;And the question and answer interactive log for not passing through at least one of words art analysis, business diagnosis and question and answer analysis is marked to be used for subsequent artefacts' quality inspection.
Description
Technical field
The present invention relates to the information processing technologies, the more particularly, to processing method of question and answer interactive log, apparatus and system.
Background technique
Customer service system is more universal with the development of e-commerce.Client can inquire interested phase by customer service system
Close information, transacting business etc..For example, user can understand information relevant to commodity, consultation service by customer service system.
Customer service system can store user's question sentence of user's inquiry and answer of attending a banquet accordingly in use, in these
Hold and saves as quality detecting data so as to subsequent examination, to guarantee customer satisfaction.The prior art is substantially by manually inspecting by random samples
Mode operated.
Quality inspection personnel is surveyed sample by the interaction data attended a banquet to phone or online text, according to quality inspection personnel to industry
The understanding of business judges whether the problem of contact staff answers be correct, and whether answer has inappropriate tone.If being related to user
The problem of account information or basic service, needs to check in systems to judge that whether correct client answers.For some
More complicated question and answer need to check example document to determine whether correct.
Above-mentioned artificial sampling observation mode is time-consuming, laborious, and quality inspection efficiency is low with sampling observation covering surface, and is difficult to find that short slab adds rapidly
To improve.
Summary of the invention
A brief summary of one or more aspects is given below to provide to the basic comprehension in terms of these.This general introduction is not
The extensive overview of all aspects contemplated, and be both not intended to identify critical or decisive element in all aspects also non-
Attempt to define the range in terms of any or all.Its unique purpose is to provide the one of one or more aspects in simplified form
A little concepts are with the sequence for more detailed description given later.
According to an aspect of the present invention, a kind of processing method for question and answer interactive log, question and answer interaction day are provided
Will includes user's question sentence and answer of attending a banquet accordingly, which includes:
Words art analysis is executed to the question and answer interactive log, prohibits language and/or negative feelings to determine that the answer of attending a banquet whether there is
Sense;
Using CRM database to the question and answer interactive log execute business diagnosis, with judge user's question sentence whether with user
The business handled matches and whether the answer of attending a banquet is consistent with the CRM data of extraction;
It executes question and answer to the question and answer interactive log using question and answer knowledge base to analyze, to determine whether the answer of attending a banquet is correct;
And
To not by the words art analysis, business diagnosis and question and answer analysis at least one of question and answer interactive log be marked
To be used for subsequent artefacts' quality inspection.
According to another aspect of the present invention, a kind of processing unit for question and answer interactive log, question and answer interaction are provided
Log includes user's question sentence and answer of attending a banquet accordingly, which includes:
Art analytical unit is talked about, for executing words art analysis to the question and answer interactive log, to determine whether the answer of attending a banquet deposits
Prohibiting language and/or negative emotion;
Business diagnosis unit, for executing business diagnosis to the question and answer interactive log using CRM database, to judge the use
Whether whether family question sentence consistent with the CRM data of extraction with the business matching and the answer of attending a banquet that user handled;
Question and answer analytical unit is used to execute question and answer to the question and answer interactive log using question and answer knowledge base and analyze, is somebody's turn to do with determining
Whether answer of attending a banquet is correct;And
Marking unit, for handing over the question and answer for not passing through at least one of words art analysis, business diagnosis and question and answer analysis
Mutual log is marked for subsequent artefacts' quality inspection.
In accordance with a further aspect of the present invention, a kind of processing system for question and answer interactive log, question and answer interaction are provided
Log includes user's question sentence and answer of attending a banquet accordingly, which includes:
CRM database, for storing CRM data associated with client;
Question and answer knowledge base asks for storing standard and asks associated model answer with each standard;And
Foregoing processing unit.
Processing method according to the present invention for question and answer interactive log can automatically hand over undesirable question and answer
Mutual log is marked, and plays the role of preliminary examination, for artificial further quality inspection.By this method, first is that substantially increasing quality inspection
Efficiency, greatly alleviate quality inspection personnel manually from magnanimity interactive log quality inspection workload.On the other hand, due to quality inspection personnel
Quality inspection can only be carried out by the way of sampling observation, cause quality inspection result that can not react truth completely.According to the solution of the present invention, right
The case where all question and answer interactive logs are all handled, avoid missing inspection.
Detailed description of the invention
After the detailed description for reading embodiment of the disclosure in conjunction with the following drawings, it better understood when of the invention
Features described above and advantage.In the accompanying drawings, each component is not necessarily drawn to scale, and has similar correlation properties or feature
Component may have same or similar appended drawing reference.
Fig. 1 is to show the flow chart of the processing method for question and answer interactive log according to an aspect of the present invention;
Fig. 2 is to show the block diagram of the processing unit for question and answer interactive log according to an aspect of the present invention;
Fig. 3 is to show the block diagram of the business diagnosis unit according to an embodiment;
Fig. 4 is to show the block diagram of the question and answer analytical unit according to an embodiment;And
Fig. 5 is to show the block diagram of the processing system for question and answer interactive log according to an aspect of the present invention.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.Note that below in conjunction with attached drawing and specifically real
The aspects for applying example description is merely exemplary, and is understood not to carry out any restrictions to protection scope of the present invention.
Customer service system can generate a large amount of question and answer interactive log in use, describe user's question sentence and corresponding
It attends a banquet answer.According to an aspect of the present invention, a kind of processing method for question and answer interactive log is provided, it can be automatically right
Undesirable question and answer interactive log is marked, and plays the role of preliminary examination, for artificial further quality inspection.By this method,
First is that substantially increase quality inspection efficiency, greatly alleviate quality inspection personnel manually from magnanimity interactive log quality inspection workload.Separately
On the one hand, since quality inspection personnel can only carry out quality inspection by the way of sampling observation, cause quality inspection result that can not react truth completely.
According to the solution of the present invention, the case where all question and answer interactive logs all being handled, avoid missing inspection.
Fig. 1 is to show the flow chart of the processing method 100 for question and answer interactive log according to an aspect of the present invention.
As shown in Figure 1, processing method 100 may include following steps:
Step 101: pretreatment is executed to question and answer interactive log.
Interactive log can be phonetic matrix.For example, customer service system may be phone customer service system, resulting question and answer
Interactive log is the form of voice.
In the case where phonetic matrix, the question and answer interactive log of speech form can be converted to text formatting first by pretreatment
Question and answer interactive log.As example, ASR (Automatic Speech Recognition, automatic speech recognition) can be passed through
Execute the conversion of voice to text.
More preferably, pretreatment, which may also include, is filtered question and answer data meaningless in question and answer interactive log.
Step 102: question and answer interactive log being executed and talks about art analysis, be whether there is with determining answer of attending a banquet and prohibited language and/or bear
Face emotion.
The judgement for prohibiting language can be realized by prohibiting words and phrases library.In one example, it is provided with and prohibits words and phrases library, wherein having included quilt
It is any inappropriate including for example uncivil term, the word for not meeting state's laws regulation etc. and answer labeled as the word for prohibiting language
Forbidden word.Specifically, can be marked based on whether there is in the answer of attending a banquet for prohibiting words and phrases library searching question and answer interactive log
It is denoted as the word for prohibiting language, and if it exists, do not pass through words art analysis then.
Other than prohibiting language judgement, words art analysis can also judge to attend a banquet answer with the presence or absence of negative emotion, be judged with this
The attitude of customer service.Sentiment analysis is measured generally in terms of two, Sentiment orientation direction and Sentiment orientation degree.
Sentiment orientation direction is also referred to as feeling polarities, it can be understood as user expresses certain object the state that itself viewpoint is held
Degree is to support, oppose, is neutral, i.e. usually signified positive emotion, negative emotion, neutral emotion.Sentiment orientation degree refers to main body
Degree of strength when positive emotion or negative emotion is expressed object, and different emotion degree is often by different emotion words
Or emotion tone etc. embodies.
In one example, the judgement of negative emotion can be realized by sentiment dictionary.Such as, it is possible to provide in love sense dictionary,
It has wherein included and has been marked as the word with negative emotion.Specifically, question and answer interactive log can be retrieved based on sentiment dictionary
Answer of attending a banquet in the presence or absence of being marked as the text with negative emotion, and if it exists, do not pass through words art analysis then.
In another example, the judgement of negative emotion can be realized by emotion classifiers.In this embodiment, first may be used
Method based on machine learning trains emotion classifiers as training set using Large Scale Corpus.The training side of classifier
Method is known technology, and details are not described herein.
Above-mentioned emotional semantic classification and taboo language judges can be to be performed using word, sentence or paragraph as unit.A paragraph piece
Chapter grade sentiment analysis carries out tendentiousness judgement primarily directed to some theme or event, generally requires the emotion for constructing corresponding event
Dictionary.The sentiment analysis of Sentence-level is obtained by calculating the average value for all emotion words for including mostly in sentence.
Both talking about art analysis can individually include prohibiting language judgement or Judgment by emotion, can also include simultaneously.For example, can be first
It whether there is the word for being marked as prohibiting language based on prohibiting in answer of attending a banquet described in words and phrases library searching question and answer interactive log, if it exists
Then not by words art analysis, otherwise continue to retrieve to whether there is in the answer of attending a banquet of question and answer interactive log based on sentiment dictionary to be marked
It is denoted as the text with negative emotion, is then otherwise analyzed if it exists by the words art not by words art analysis.It is such first to prohibit language
The mode for judging again Judgment by emotion, both ensure that the high efficiency of analysis, and had also ensured the high-accuracy of analysis.
Step 103: using CRM database to question and answer interactive log execute business diagnosis, with judge user's question sentence whether with
The business that user handled matches and whether answer of attending a banquet is consistent with the CRM data of extraction.
CRM number is stored in CRM (Customer relationship management, customer relation management) database
According to CRM data has recorded business datum relevant to client.User is in inquiry and itself personal relevant individual business problem
When, such as oneself what set meal, inquiry into balance etc. being had subscribed, customer service needs to inquire CRM database to answer.Business diagnosis master
The individual business problem relevant to user itself to propose for user is analyzed.
In one example, user's question sentence in question and answer interactive log can be segmented first.It can be used in the present invention
Any suitable segmentation methods segment user's question sentence.Common segmentation methods may include character match method, understanding method, system
Meter method etc..
After participle, CRM data relevant to word segmentation result can be extracted from CRM database.In one example, can pass through by
Word segmentation result is matched with CRM data in CRM database, to obtain CRM data relevant to word segmentation result.Then, from CRM
CRM data relevant to word segmentation result is extracted in database, and judges whether the CRM data extracted and answer of attending a banquet are consistent,
As unanimously, then passed through business diagnosis.
Such as user's question sentence may is that inquiry into balance;
Answer of attending a banquet accordingly may is that 48 yuan.
According to examples detailed above, it is corresponding that the user balance this CRM field is transferred from CRM database based on inquiry into balance
Data: 45 yuan.
At this point, then business diagnosis does not pass through since 48 yuan of answer of attending a banquet inconsistent with 45 yuan of CRM data.
When extracting CRM data, it is also possible to extract failure, such as user's question sentence is not one relevant to user identity
Traffic issues, at this point it is possible to Null be returned to, to indicate that business diagnosis is invalid.
Step 104: question and answer being executed to question and answer interactive log using question and answer knowledge base and are analyzed, whether just answer is attended a banquet with determination
Really.
A large amount of professional knowledge point is stored in question and answer knowledge base in the form of problem-answer knowledge point." standard is asked " is
For indicating the text of some knowledge point, main target is that expression is clear, convenient for safeguarding.For example, " rate of CRBT " are exactly table
Description is asked up to clearly standard.Here " asking " narrowly should not be interpreted as " inquiring ", and should broadly understand that one is " defeated
Enter ", being somebody's turn to do " input " has corresponding " output ".For example, for the semantics recognition for control system, the finger of user
It enables, such as " opening radio " should also be understood to be one " asking ", corresponding at this time " answering " can be for executing phase
The calling for the control program that should be controlled.
In one example, it may be determined that whether there is in question and answer knowledge base and match with user's question sentence in question and answer interactive log
Standard ask.This judgement can be executed by Semantic Similarity Measurement.For example, user's question sentence and the standard in knowledge base can be asked
Semantic Similarity Measurement is executed, the standard with highest semantic similarity is asked and is asked as matching standard.
Arithmetic of Semantic Similarity refers to through certain method, to calculate the similarity degree between two different words and expressions.
The semantic similarity degree between sentence would generally be measured with a percentage.
Common similarity of character string algorithm includes editing distance algorithm (EditDistance), n-gram algorithm,
JaroWinkler algorithm and Soundex algorithm etc..
The similarity problem of two sentences can be attributed to and be converted to one of sentence character string by editing distance algorithm
Another cost to be paid of sentence character string.The cost of conversion is higher, illustrates that the similarity of two character strings is lower.Usually
The transform mode that can choose includes insertion, replacement and deletion.
N-Gram algorithm is based on such a hypothesis: the appearance of n-th of word and front n-1 i.e. in character string
A word is related, and all uncorrelated to other any words, and the probability that entire character string occurs is exactly multiplying for the probability that each word occurs
Product.N-gram itself represents length in target string also as the substring of n, and citing, " ARM " is a 3- in " ARMY "
gram.In two character strings, when identical n-gram is more, two word strings will be considered even more like.
Jaro Winkler is then that n-gram algorithm more goes a step further.By the unmatched part in n-gram simultaneously into
Gone transposition the considerations of, make it possible to obtain more accurate similarity degree.JaroWinkler is comparing two feelings compared with short character strings
Under condition, good result can be obtained.
For example, user's question sentence can be with are as follows: I wants to ask lower current 88 yuan of 4G set meals include how many flow.
Matched standard is asked can be with are as follows: 4G set meal flow.
Matched standard is asked if it exists, it is determined that the corresponding model answer asked of matched standard and question and answer interactive log
In accordingly answer of attending a banquet it is whether consistent, analyzed if matching by question and answer, otherwise do not pass through question and answer analysis.
For example, asking that " 4G set meal flow " corresponding model answer is that a function of call parameters " 88 yuan " is returned with standard
Return value, such as the flow of 2 G.At this time can be by it compared with answer of attending a banquet pair, i.e., knowable answer correctness of attending a banquet.
Question and answer analysis be directed to it is mostly be general traffic issues, be directed to inquiry into balance this kind and the close phase of user identity
When the traffic issues of pass carry out question and answer analysis, the knowledge point of this personalization, therefore question and answer are not had in question and answer knowledge base generally
The result of analysis is null, indicates that question and answer analysis is invalid.
In some embodiments, question and answer analysis may also be combined with long text parser, can be by by long text analysis method
It determines that user's question sentence includes M problem, and M sub- answers is divided by the answer that will attend a banquet of long text analysis method, in this base
The enterprising hand-manipulating of needle of plinth carries out question and answer analysis to each problem and its corresponding sub- answer respectively.Only when all problems and its corresponding
When sub- answer is all analyzed by question and answer, it just can determine that answer of attending a banquet is correct.
Step 105: to not by words art analysis, business diagnosis and question and answer analysis at least one of question and answer interactive log into
Line flag is to be used for subsequent artefacts' quality inspection.
By to not by words art analysis, business diagnosis and question and answer analysis at least one of question and answer interactive log mark
Note can make manually these question and answer interactive logs for being marked with of only quality inspection, to save manpower.
In one example, customer service information can be extracted from labeled question and answer interactive log, the customer service information can wrap
Include: temporal information, personal information, labeled type are (that is: not by the analysis of words art, not by business diagnosis and/or not by asking
Answer analysis), labeled number etc., corresponding contact staff is targetedly trained or is punished so as to subsequent.
Art analysis, business diagnosis and question and answer analysis can select therein any one according to the user's choice if above-mentioned
Kind is executed for two or three.Alternatively, the relationship that can be analyzed by presetting user's question sentence with quality inspection, thus subsequent receipts
Real-time quality inspection scheme is generated when to specific user's question sentence.For example, being relevant to user identity when receiving user's question sentence
Property traffic issues when, can execute business diagnosis, and be the general traffic issues unrelated with user identity in user's question sentence
When, question and answer analysis can be executed.
Although for simplify explain the above method is illustrated to and is described as a series of actions, it should be understood that and understand,
The order that these methods are not acted is limited, because according to one or more embodiments, some movements can occur in different order
And/or with from it is depicted and described herein or herein it is not shown and describe but it will be appreciated by those skilled in the art that other
Movement concomitantly occurs.
Fig. 2 is to show the block diagram of the processing unit 200 for question and answer interactive log according to an aspect of the present invention.Place
Reason device 200 may include pretreatment unit 201, words art analytical unit 202, business diagnosis unit 203, question and answer analytical unit 204,
Marking unit 205.
The question and answer interactive log of phonetic matrix can be converted to the question and answer interactive log of text formatting simultaneously by pretreatment unit 201
Filter meaningless question and answer data.
Words art analytical unit 202 can be used for executing question and answer interactive log words art analysis, to determine whether answer of attending a banquet deposits
Prohibiting language and/or negative emotion.In one example, words art analytical unit 202 can be based on taboo words and phrases library searching question and answer interactive log
Answer of attending a banquet in the presence or absence of be marked as prohibit language word, and if it exists, then by words art analysis.In another example,
Words art analytical unit 202 can be retrieved to whether there is in the answer of attending a banquet of question and answer interactive log based on sentiment dictionary and be marked as having
The text of negative emotion, and if it exists, do not pass through words art analysis then.In another example, words art analytical unit 202 can be based on language
Expect that the emotion classifiers built-up as training set carry out emotional semantic classification to the answer of attending a banquet of question and answer interactive log, when classification is tied
When fruit is negative emotion, then do not pass through words art analysis.
Above-mentioned words art analytical unit 202 can execute words art analysis using word, sentence or paragraph as unit.
Business diagnosis unit 203 executes business diagnosis to question and answer interactive log using CRM database, to judge user's question sentence
It is whether consistent with the CRM data of extraction with the business matching and answer of attending a banquet that user handled.
In one example, as shown in figure 3, business diagnosis unit 203 may include participle unit 2031, CRM data acquisition list
Member 2032 and comparing unit 2033.Participle unit 2031 can segment user's question sentence in question and answer interactive log, CRM
Data capture unit 2032 can extract CRM data relevant to word segmentation result from CRM database, and comparing unit 2033 can be sentenced
Whether the disconnected CRM data extracted and answer of attending a banquet are consistent, such as consistent, then pass through business diagnosis.
Question and answer analytical unit 204 can execute question and answer to question and answer interactive log using question and answer knowledge base and analyze, and be attended a banquet with determination
Whether answer is correct.
Specifically, question and answer analytical unit 204 can determine in question and answer knowledge base with the presence or absence of in question and answer interactive log
The standard that user's question sentence matches is asked, and if it exists, then determines the corresponding model answer asked of the matched standard of institute and the question and answer
Whether the answer of attending a banquet in interactive log is consistent, is analyzed if matching by question and answer.
In one example, as shown in figure 4, question and answer analytical unit 204 may include Semantic Similarity Measurement unit 2041 with logical
Semantic Similarity Measurement is crossed to determine the matching of user's question sentence.In another example, question and answer analytical unit 204 may include long text
Analytical unit 2042 is to be divided into M problem for user's question sentence, and the answer that will attend a banquet is divided into M sub- answers.
Marking unit 205 can be interactive to the question and answer not by least one of the analysis of words art, business diagnosis and question and answer analysis
Log is marked for subsequent artefacts' quality inspection.
In one embodiment, which may also include extraction unit to mention from labeled question and answer interactive log
Take customer service information.
According to another aspect of the present invention, a kind of processing system for question and answer interactive log is additionally provided, such as Fig. 5 institute
Show.Processing system 500 may include CRM database 510, question and answer knowledge base 520 and processing unit 530.Here processing unit
530 can have similar structure with previously described processing unit 200.
Processing unit 530 in Fig. 5 illustrates only business diagnosis unit 531 and question and answer analytical unit 532.Business diagnosis list
Member 531 can execute business diagnosis to question and answer interactive log using CRM database, to judge whether user's question sentence was handled with user
Business matching and answer of attending a banquet it is whether consistent with the CRM data of extraction.Question and answer analytical unit 532 can utilize question and answer knowledge base pair
Question and answer interactive log executes question and answer analysis, to determine whether answer of attending a banquet is correct.
Offer is to make any person skilled in the art all and can make or use this public affairs to the previous description of the disclosure
It opens.The various modifications of the disclosure all will be apparent for a person skilled in the art, and as defined herein general
Suitable principle can be applied to other variants without departing from the spirit or scope of the disclosure.The disclosure is not intended to be limited as a result,
Due to example described herein and design, but should be awarded and principle disclosed herein and novel features phase one
The widest scope of cause.
Claims (25)
1. a kind of processing method for question and answer interactive log, the question and answer interactive log includes user's question sentence and corresponding seat
Seat answer, the treating method comprises:
Words art analysis is executed to the question and answer interactive log, with answer of attending a banquet described in determination with the presence or absence of taboo language and/or negative feelings
Sense;
Using CRM database to the question and answer interactive log execute business diagnosis, with judge user's question sentence whether with user
The business handled matches and whether the answer of attending a banquet is consistent with the CRM data of extraction;
Question and answer are executed to the question and answer interactive log using question and answer knowledge base to analyze, it is whether correct with answer of attending a banquet described in determination;
And
To not by it is described words art analysis, business diagnosis and question and answer analysis at least one of question and answer interactive log be marked with
For subsequent artefacts' quality inspection.
2. processing method as described in claim 1, which is characterized in that the execution words art, which is analyzed, includes:
It whether there is the word for being marked as prohibiting language based on prohibiting in answer of attending a banquet described in question and answer interactive log described in words and phrases library searching
Language, and if it exists, do not analyzed by the words art then.
3. processing method as described in claim 1, which is characterized in that the execution words art, which is analyzed, includes:
It is retrieved based on sentiment dictionary negative with the presence or absence of being marked as having in answer of attending a banquet described in the question and answer interactive log
The text of emotion, and if it exists, do not analyzed by the words art then.
4. processing method as described in claim 1, which is characterized in that the execution words art, which is analyzed, includes:
Based on using corpus as the built-up emotion classifiers of training set to answer of attending a banquet described in the question and answer interactive log
Emotional semantic classification is carried out, when classification results are negative emotion, is not then analyzed by the words art.
5. processing method as described in claim 3 or 4, which is characterized in that the words art analysis is with word, sentence or section
It falls and is performed for unit.
6. processing method as described in claim 1, which is characterized in that the execution words art, which is analyzed, includes:
It whether there is the word for being marked as prohibiting language based on prohibiting in answer of attending a banquet described in question and answer interactive log described in words and phrases library searching
Language is not analyzed then by the words art if it exists, otherwise continues to be retrieved based on sentiment dictionary described in the question and answer interactive log
It attends a banquet with the presence or absence of the text for being marked as that there is negative emotion in answer, if it exists then not by words art analysis, otherwise
It is analyzed by the words art.
7. processing method as described in claim 1, which is characterized in that the execution business diagnosis includes:
User's question sentence in the question and answer interactive log is segmented;
CRM data relevant to word segmentation result is extracted from CRM database, and
Judge whether the CRM data extracted and the answer of attending a banquet are consistent, it is such as consistent, then pass through the business diagnosis.
8. processing method as described in claim 1, which is characterized in that the execution question and answer, which are analyzed, includes:
It determines in the question and answer knowledge base with the presence or absence of the mark to match with user's question sentence in the question and answer interactive log
Standard is asked;
If it exists, it is determined that the corresponding model answer asked of matched standard attend a banquet with described in the question and answer interactive log
Whether answer is consistent, is analyzed if matching by the question and answer.
9. processing method as claimed in claim 8, which is characterized in that the determination of user's question matching passes through semantic similar
Degree calculates to execute.
10. processing method as claimed in claim 9, which is characterized in that further include: by described in the determination of long text analysis method
User's question sentence includes M problem, and the answer of attending a banquet is divided into M sub- answers by long text analysis method.
11. processing method as described in claim 1, which is characterized in that further include:
The question and answer interactive log of phonetic matrix is converted into the question and answer interactive log of text formatting and filters meaningless question and answer number
According to.
12. processing method as described in claim 1, which is characterized in that further include:
Customer service information is extracted from labeled question and answer interactive log.
13. a kind of processing unit for question and answer interactive log, the question and answer interactive log includes user's question sentence and corresponding
It attends a banquet answer, the processing unit includes:
Art analytical unit is talked about, for executing words art analysis to the question and answer interactive log, whether answer is deposited to attend a banquet described in determination
Prohibiting language and/or negative emotion;
Business diagnosis unit, for executing business diagnosis to the question and answer interactive log using CRM database, to judge the use
Whether whether family question sentence consistent with the CRM data of extraction with the business matching and the answer of attending a banquet that user handled;
Question and answer analytical unit is analyzed for executing question and answer to the question and answer interactive log using question and answer knowledge base, described in determination
Whether answer of attending a banquet is correct;And
Marking unit, for interactive to the question and answer for not passing through at least one of the words art analysis, business diagnosis and question and answer analysis
Log is marked for subsequent artefacts' quality inspection.
14. processing unit as claimed in claim 13, which is characterized in that the words art analytical unit is based on prohibiting words and phrases library searching
With the presence or absence of the word for being marked as taboo language in the answer of attending a banquet of the question and answer interactive log, and if it exists, do not pass through institute then
State words art analysis.
15. processing unit as claimed in claim 13, which is characterized in that the words art analytical unit is retrieved based on sentiment dictionary
With the presence or absence of the text for being marked as that there is negative emotion in the answer of attending a banquet of the question and answer interactive log, and if it exists, then
It is not analyzed by the words art.
16. processing unit as claimed in claim 13, which is characterized in that the words art analytical unit is based on using corpus as instruction
Practice the emotion classifiers for collecting built-up and emotional semantic classification is carried out to answer of attending a banquet described in the question and answer interactive log, when classification is tied
When fruit is negative emotion, then do not analyzed by the words art.
17. the processing unit as described in claim 15 or 16, which is characterized in that the words art analytical unit is with word, sentence
Son or paragraph are that unit is performed words art analysis.
18. processing unit as claimed in claim 13, which is characterized in that the words art analytical unit is based on prohibiting words and phrases library searching
With the presence or absence of the word for being marked as taboo language in the answer of attending a banquet of the question and answer interactive log, if it exists then by described
Art analysis is talked about, otherwise continues to retrieve based on sentiment dictionary and be attended a banquet in answer described in the question and answer interactive log with the presence or absence of being marked
It is denoted as the text with negative emotion, is not analyzed if it exists by the words art then, is otherwise analyzed by the words art.
19. processing unit as claimed in claim 13, which is characterized in that the business diagnosis unit includes:
Participle unit segments user's question sentence in the question and answer interactive log;
CRM data acquiring unit extracts CRM data relevant to word segmentation result from CRM database, and
Comparing unit judges whether the CRM data extracted and the answer of attending a banquet are consistent, such as consistent, then passes through the business point
Analysis.
20. processing unit as claimed in claim 13, which is characterized in that the question and answer analytical unit determines the question and answer knowledge
It is asked in library with the presence or absence of the standard to match with user's question sentence in the question and answer interactive log;If it exists, it is determined that institute
Whether the corresponding model answer that matched standard is asked and the answer of attending a banquet in the question and answer interactive log are consistent, if matching
Then analyzed by the question and answer.
21. processing unit as claimed in claim 20, which is characterized in that the question and answer analytical unit includes semantic similarity meter
Unit is calculated to determine the matching of user's question sentence by Semantic Similarity Measurement.
22. processing unit as claimed in claim 21, which is characterized in that the question and answer analytical unit further includes long text analysis
The answer of attending a banquet is divided into M sub- answers so that user's question sentence is divided into M problem by unit.
23. processing unit as claimed in claim 13, which is characterized in that further include:
Pretreatment unit, for the question and answer interactive log of phonetic matrix to be converted to the question and answer interactive log of text formatting and is filtered
Meaningless question and answer data.
24. processing unit as claimed in claim 13, which is characterized in that further include:
Extraction unit, for extracting customer service information from labeled question and answer interactive log.
25. a kind of processing system for question and answer interactive log, the question and answer interactive log includes user's question sentence and corresponding
It attends a banquet answer, the processing system includes:
CRM database, for storing CRM data associated with client;
Question and answer knowledge base asks for storing standard and asks associated model answer with each standard;And
Processing unit as described in any one of claim 13-24.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611061635.8A CN106776832B (en) | 2016-11-25 | 2016-11-25 | Processing method, apparatus and system for question and answer interactive log |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611061635.8A CN106776832B (en) | 2016-11-25 | 2016-11-25 | Processing method, apparatus and system for question and answer interactive log |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106776832A CN106776832A (en) | 2017-05-31 |
CN106776832B true CN106776832B (en) | 2019-07-19 |
Family
ID=58911129
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611061635.8A Active CN106776832B (en) | 2016-11-25 | 2016-11-25 | Processing method, apparatus and system for question and answer interactive log |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106776832B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107766560B (en) * | 2017-11-03 | 2020-09-15 | 广州杰赛科技股份有限公司 | Method and system for evaluating customer service flow |
CN108153732B (en) * | 2017-12-25 | 2021-08-03 | 浙江讯飞智能科技有限公司 | Examination method and device for interrogation notes |
CN108491388B (en) * | 2018-03-22 | 2021-02-23 | 平安科技(深圳)有限公司 | Data set acquisition method, classification method, device, equipment and storage medium |
CN108712581A (en) * | 2018-06-07 | 2018-10-26 | 上海航动科技有限公司 | A kind of traffic method of skill training, device, equipment and system |
CN112086092A (en) * | 2019-06-14 | 2020-12-15 | 广东技术师范大学 | Intelligent extraction method of dialect based on emotion analysis |
CN111309881A (en) * | 2020-02-11 | 2020-06-19 | 深圳壹账通智能科技有限公司 | Method and device for processing unknown questions in intelligent question answering, computer equipment and medium |
CN111488439B (en) * | 2020-03-18 | 2023-04-18 | 杭州迪普科技股份有限公司 | System and method for saving and analyzing log data |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101076184A (en) * | 2006-07-31 | 2007-11-21 | 腾讯科技(深圳)有限公司 | Method and system for realizing automatic reply |
CN102291617A (en) * | 2011-09-03 | 2011-12-21 | 四川公用信息产业有限责任公司 | End-to-end fault diagnosing and positioning platform of IPTV (Internet Protocol Television) business |
CN104679910A (en) * | 2015-03-25 | 2015-06-03 | 北京智齿博创科技有限公司 | Intelligent answering method and system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006085661A1 (en) * | 2005-02-08 | 2006-08-17 | Nec Corporation | Question answering data edition device, question answering data edition method, and question answering data edition program |
-
2016
- 2016-11-25 CN CN201611061635.8A patent/CN106776832B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101076184A (en) * | 2006-07-31 | 2007-11-21 | 腾讯科技(深圳)有限公司 | Method and system for realizing automatic reply |
CN102291617A (en) * | 2011-09-03 | 2011-12-21 | 四川公用信息产业有限责任公司 | End-to-end fault diagnosing and positioning platform of IPTV (Internet Protocol Television) business |
CN104679910A (en) * | 2015-03-25 | 2015-06-03 | 北京智齿博创科技有限公司 | Intelligent answering method and system |
Also Published As
Publication number | Publication date |
---|---|
CN106776832A (en) | 2017-05-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106776832B (en) | Processing method, apparatus and system for question and answer interactive log | |
CN106295807B (en) | A kind of method and device of information processing | |
CN109151218A (en) | Call voice quality detecting method, device, computer equipment and storage medium | |
US9501467B2 (en) | Systems, methods, software and interfaces for entity extraction and resolution and tagging | |
Cummins et al. | Enhancing speech-based depression detection through gender dependent vowel-level formant features | |
CN108549637A (en) | Method for recognizing semantics, device based on phonetic and interactive system | |
CN105930452A (en) | Smart answering method capable of identifying natural language | |
CN112765974B (en) | Service assistance method, electronic equipment and readable storage medium | |
Levitan et al. | Combining Acoustic-Prosodic, Lexical, and Phonotactic Features for Automatic Deception Detection. | |
Garnier-Rizet et al. | CallSurf: Automatic Transcription, Indexing and Structuration of Call Center Conversational Speech for Knowledge Extraction and Query by Content. | |
Zvarevashe et al. | Recognition of speech emotion using custom 2D-convolution neural network deep learning algorithm | |
CN110765776B (en) | Method and device for generating return visit labeling sample data | |
Kopparapu | Non-linguistic analysis of call center conversations | |
Verbree et al. | Dialogue-act tagging using smart feature selection; results on multiple corpora | |
Wagner et al. | Applying cooperative machine learning to speed up the annotation of social signals in large multi-modal corpora | |
Sanabria et al. | The edinburgh international accents of english corpus: Towards the democratization of english asr | |
Koudounas et al. | Italic: An italian intent classification dataset | |
CN111916110A (en) | Voice quality inspection method and device | |
Danner et al. | Advancing Mental Health Diagnostics: GPT-Based Method for Depression Detection | |
CN115063155B (en) | Data labeling method, device, computer equipment and storage medium | |
Braunger et al. | A comparative analysis of crowdsourced natural language corpora for spoken dialog systems | |
CN105786929A (en) | Information monitoring method and device | |
Durrani et al. | Transfer learning based speech affect recognition in Urdu | |
CN113642335A (en) | Language compliance checking method, device, equipment and medium for double-recording scene of bank | |
Mishra et al. | A review paper on voice analytics |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Processing method, device and system for Q&A interactive log Effective date of registration: 20230223 Granted publication date: 20190719 Pledgee: China Construction Bank Corporation Shanghai No.5 Sub-branch Pledgor: SHANGHAI XIAOI ROBOT TECHNOLOGY Co.,Ltd. Registration number: Y2023980033272 |
|
PE01 | Entry into force of the registration of the contract for pledge of patent right |