CN106503744A - Input expression in chat process carries out the method and device of automatic error-correcting - Google Patents
Input expression in chat process carries out the method and device of automatic error-correcting Download PDFInfo
- Publication number
- CN106503744A CN106503744A CN201610943311.0A CN201610943311A CN106503744A CN 106503744 A CN106503744 A CN 106503744A CN 201610943311 A CN201610943311 A CN 201610943311A CN 106503744 A CN106503744 A CN 106503744A
- Authority
- CN
- China
- Prior art keywords
- expression
- input
- chat
- detected
- input expression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/40—Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
Abstract
The present invention provide to chat process in input expression carry out the method and device of automatic error-correcting, by gathering the chat attribute information of training sample;Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;Grader is trained according to characteristic vector, obtain error detection model and according to error detection model, judge whether input expression to be detected is accurate, if not, error correction is carried out to input expression to be detected then, solve due to wrong choosing expression or misread expression meanings and cause the inaccurate technical problem of input expression, on the one hand by using error detection model to chat process in input expression realize automatic error detection, on the other hand inaccurate input expression realizes automatic error-correcting.
Description
Technical field
The present invention relates to communication technical field, and in particular to a kind of to chat process in input expression carry out automatic error-correcting
Method and device.
Background technology
With IM(Instant Messenger, instant messaging)Application, Blog(Blog)And SMS(Short
Messaging Service, Short Message Service)That applied popularizes, and user has depended on these that there is information further
The application of transmission-receiving function carries out exchange to each other and contact.
User, is generally required to be input into one to increase the interest of input content when exchanging using above-mentioned application
Expressed one's feelings to express particular meaning, or abundant input content a bit.However, an application has usually contained tens even hundreds of tables
Feelings are selected for user, therefore in chat process input expression may be caused inaccurate because of wrong choosing expression or misunderstanding expression meanings
Really.For the problem, the invention provides a kind of to chat process in input expression carry out the method and device of automatic error-correcting.
Content of the invention
The invention provides a kind of to chat process in input expression carry out the method and device of automatic error-correcting, to solve
As wrong choosing expression or misunderstanding expression meanings cause the inaccurate technical problem of input expression.
According to an aspect of the present invention, there is provided a kind of to chat process in input expression carry out the side of automatic error-correcting
Method, including:
The chat attribute information of collection training sample, wherein, training sample includes that input expression is wrong and input expression is error-free
Training sample by mistake;
Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;
Grader is trained according to characteristic vector, error detection model is obtained;
According to error detection model, whether accurately input expression to be detected is judged, if it is not, then entangling to input expression to be detected
Wrong.
Further, according to error detection model, judge whether input expression to be detected accurately includes:
Whether the current input of detection includes emoticon, and if so, then emoticon is expressed one's feelings as input to be detected;
The chat attribute information of collection input expression to be detected;
Based on input expression to be detected and the chat attribute information of input expression to be detected, input expression to be detected is extracted
Characteristic vector;
By the characteristic vector input error detection model of input expression to be detected, judge whether input expression to be detected is accurate.
Further, carrying out error correction to input expression to be detected includes:
In default expression data storehouse, mate the error correction table most mated with the chat attribute information of input expression to be detected
Feelings;
Error correction is expressed one's feelings and replaces input expression to be detected.
Further, in default expression data storehouse, mate the chat attribute information that expresses one's feelings with input to be detected most
The error correction expression of coupling includes:
Carry out Text Flag in advance to the expression in default expression data storehouse, obtain text expression;
Extract and the text corresponding term vector of expression, obtain text expression term vector;
Term vector corresponding with the chat attribute information of input expression to be detected is extracted, chat attribute term vector is obtained;
Calculate chat attribute term vector and text expression term vector between similarity, and by similarity highest text express one's feelings word
The corresponding expression of vector is used as the error correction expression that most mates with the chat attribute information of input expression to be detected.
Further, chat attribute information includes:
One kind or many in chatting object's relation, chatting time, chat theme, chat context and chat geographical location information
Plant combination.
According to a further aspect in the invention, there is provided a kind of to chat process in input expression carry out the dress of automatic error-correcting
Put, including:
Harvester, for gathering the chat attribute information of training sample, wherein, training sample include input expression wrong and
The faultless training sample of input expression;
Feature-vector extraction means, for the chat attribute information based on training sample and training sample, extract training sample
Characteristic vector;
Training devicess, for training grader according to characteristic vector, obtain error detection model;
Whether accurately error correction device, for according to error detection model, judging input expression to be detected, if it is not, then to be detected
Input expression carries out error correction.
Further, error correction device includes:
Detection means, for detecting current input whether comprising emoticon, if so, then using emoticon as to be detected defeated
Enter expression;
Chat attribute information harvester, for gathering the chat attribute information of input expression to be detected;
Input expressive features vector extraction device, for the chat based on input expression to be detected and input expression to be detected
Attribute information, extracts the characteristic vector of input expression to be detected;
Judgment means, for the characteristic vector input error detection model by input expression to be detected, judge input table to be detected
Whether feelings are accurate.
Further, error correction device also includes:
Coalignment, in default expression data storehouse, mating the chat attribute information that expresses one's feelings with input to be detected most
The error correction expression of coupling;
Alternative, replaces input expression to be detected for error correction is expressed one's feelings.
Further, coalignment includes:
Identity device, for carrying out Text Flag in advance to the expression in default expression data storehouse, obtains text expression;
Text expression term vector acquisition device, for extraction and the text corresponding term vector of expression, obtains text expression term vector;
Chat attribute term vector acquisition device, for extract word corresponding with the chat attribute information of input expression to be detected to
Amount, obtains chat attribute term vector;
Similarity Measure device, for calculating the similarity between chat attribute term vector and text expression term vector, and by phase
Seemingly the corresponding expression of degree highest text expression term vector is most mated as the chat attribute information that is expressed one's feelings with input to be detected
Error correction expression.
The invention has the advantages that:
The present invention provide to chat process in input expression carry out the method and device of automatic error-correcting, by gathering training sample
This chat attribute information;Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;
Grader is trained according to characteristic vector, error detection model is obtained and according to error detection model, whether is judged input expression to be detected
Accurately, if it is not, then carrying out error correction to input expression to be detected, solve due to wrong choosing expression or misread expression meanings cause defeated
Enter inaccurate technical problem of expressing one's feelings, on the one hand by using error detection model to chat process in input expression realize automatically examining
Mistake, on the other hand inaccurate input expression realize automatic error-correcting.
In addition to objects, features and advantages described above, the present invention also has other objects, features and advantages.
Below with reference to figure, the present invention is further detailed explanation.
Description of the drawings
The accompanying drawing for building the part of the application is used for providing a further understanding of the present invention, the schematic reality of the present invention
Apply example and its illustrate, for explaining the present invention, not building inappropriate limitation of the present invention.In the accompanying drawings:
Fig. 1 be the preferred embodiment of the present invention to chat process in input expression carry out the method flow diagram of automatic error-correcting;
Fig. 2 be directed to one of the preferred embodiment of the present invention simplify embodiment to chat process in input expression carry out automatically
The method flow diagram of error correction;
Fig. 3 be the preferred embodiment of the present invention to chat process in input expression carry out automatic error-correcting device structural frames
Figure.
Description of reference numerals:
10th, harvester;20th, feature-vector extraction means;30th, training devicess;40th, error correction device.
Specific embodiment
Embodiments of the invention are described in detail below in conjunction with accompanying drawing, but the present invention can be defined by the claims
Implement with the multitude of different ways for covering.
With reference to Fig. 1, the preferred embodiments of the present invention provide a kind of to chat process in input expression entangled automatically
Wrong method, including:
Step S101, gathers the chat attribute information of training sample, and wherein, training sample includes that input expression is wrong and is input into
Express one's feelings faultless training sample;
Step S102, based on the chat attribute information of training sample and training sample, extracts the characteristic vector of training sample;
Step S103, trains grader according to characteristic vector, obtains error detection model;
Whether accurately step S104, according to error detection model, judge input expression to be detected, if it is not, then to input to be detected
Expression carries out error correction.
The present invention provide to chat process in input expression carry out automatic error-correcting method, by gather training sample
Chat attribute information;Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;Root
Grader is trained according to characteristic vector, error detection model is obtained and according to error detection model, is judged whether input expression to be detected is accurate
Really, if it is not, then carrying out error correction to input expression to be detected, solving causes to be input into due to wrong choosing expression or misunderstanding expression meanings
Express one's feelings inaccurate technical problem, on the one hand by using error detection model to chat process in input expression realize examining automatically
Mistake, on the other hand inaccurate input expression realize automatic error-correcting.
Specifically, the present embodiment carries out automatic error-correcting using based on error detection model to being input into expression, dexterously by error correction
Error detection problem in journey is converted to classification problem, carries out the troublesome operation of error detection so as to simplify prior art to input expression,
Substantially increase carries out the efficiency and error detection rate of error detection to input expression, is the follow-up error correction for improving and carrying out error correction to input expression
Efficiency and error correction rate lay the foundation;Secondly the present embodiment is set up and trains error detection by using the chat attribute information of training sample
Model so that the error detection model of foundation fully combines the chat attribute information of input expression, so as to have higher error detection rate,
The error correction rate to input expression is further increased, higher intelligent error correction level is embodied.
According to daily chat, whether input expression is correct or suitable not only relevant with the implication of input expression itself,
Also related to the chat attribute information of input expression(Such as chatting object's relation, chatting time, chat theme, chat context
And chat geographical location information etc.).For example, in the chat scenario for two different chat relationships, using same defeated
Entering expression may be correct in a chat scenario wherein, and incorrect in another chat scenario.For the problem, this reality
Example is applied when training and set up error detection model, the chat attribute information of input expression is fully combined, so as to have higher error detection
Rate, further increases the error correction rate to input expression, embodies higher intelligent error correction level.
In actual implementation process, in order to improve the error detection for carrying out error detection using error detection model to input expression as far as possible
Rate, the present embodiment should choose enough training sample training graders, and be input into wrong and input expression inerrancy of expressing one's feelings
Training sample number as far as possible quite.
Alternatively, according to error detection model, judge whether input expression to be detected accurately includes:
Whether the current input of detection includes emoticon, and if so, then emoticon is expressed one's feelings as input to be detected;
The chat attribute information of collection input expression to be detected;
Based on input expression to be detected and the chat attribute information of input expression to be detected, input expression to be detected is extracted
Characteristic vector;
By the characteristic vector input error detection model of input expression to be detected, judge whether input expression to be detected is accurate.
The present embodiment is after training for carrying out the error detection model of error detection to input expression, it is possible to using the error detection mould
Type judges whether input expression to be detected is accurate.Specifically, detect whether current input includes emoticon first, if so,
Then emoticon is expressed one's feelings as input to be detected, then gathers the chat attribute information of input expression to be detected, and base
In input expression to be detected and the chat attribute information of input expression to be detected, the feature of input expression to be detected is extracted
Vector, finally by the characteristic vector input error detection model of input expression to be detected, judges whether input expression to be detected is accurate
Really.
Specifically, whether the current input of the present embodiment detection can be real-time comprising emoticon, it is also possible to be non-reality
When, namely whether can be belonged to or including emoticon with the content of real-time detection chat input, it is also possible to by default input
Amount periodically detects whether current input belongs to or comprising emoticon.In specific implementation process, the present embodiment can be with
Concrete setting start to detect whether the content that chat is input into belongs to or including emoticon, for example may be used when input quantity is how many characters
Input quantity is arranged as 5 characters, namely when being often input into 5 characters, judge whether 5 characters belong to or including emoticon
Number.It should be noted that only just expressing one's feelings as input to be detected when current input is detected comprising emoticon
And whether accurately to judge which, and the input that the chat attribute information of input expression to be detected specifically can be to be detected is expressed one's feelings
One or more group in chatting object's relation, chatting time, chat theme, chat context and chat geographical location information
Close.
The present embodiment by the error detection problem in error correction procedure is converted to classification problem, so as to simplify prior art pair
Input expression carries out the troublesome operation of error detection, and substantially increase carries out the efficiency and error detection rate of error detection to input expression, is follow-up
Improve and the error correction efficiency and error correction rate that input expression carries out error correction is laid the foundation.And the present embodiment is using the error detection mould for training
The error detection to input expression to be detected realized by type, fully combines the chat attribute information of input expression to be detected, significantly
Improve the error detection rate to input expression to be detected.
Alternatively, carrying out error correction to input expression to be detected includes:
In default expression data storehouse, mate the error correction table most mated with the chat attribute information of input expression to be detected
Feelings;
Error correction is expressed one's feelings and replaces input expression to be detected.
The present embodiment according to the error detection model inspection that trains go out input expression inaccurate when, inaccurate input table
Feelings carry out automatic error-correcting.Specifically, first in default expression data storehouse, mate the chat category with input expression to be detected
Property the error correction expression most mated of information, then error correction is expressed one's feelings and replaces input expression to be detected.The present embodiment passes through default
Expression data storehouse in, the error correction expression that Auto-matching is most mated with the chat attribute information of input expression to be detected, and will
Error correction expression replaces input expression to be detected, it is achieved that inaccurate input expression carries out automatic error-correcting, without the need for manually joining
With error correction efficiency is high;Additionally, the present embodiment is most mated with which with reference to the chat attribute information coupling of input expression to be detected
Error correction is expressed one's feelings, and intelligence degree is high.
Alternatively, in default expression data storehouse, mate with the chat attribute information of input expression to be detected most
The error correction expression that matches somebody with somebody includes:
Carry out Text Flag in advance to the expression in default expression data storehouse, obtain text expression;
Extract and the text corresponding term vector of expression, obtain text expression term vector;
Term vector corresponding with the chat attribute information of input expression to be detected is extracted, chat attribute term vector is obtained;
Calculate chat attribute term vector and with text express one's feelings term vector between similarity, and by similarity highest text express one's feelings
The corresponding expression of term vector is used as the error correction expression that most mates with the chat attribute information of input expression to be detected.
Due to the chat attribute information of input expression to be detected potentially include multiple, therefore in order in default expression data
The error correction expression that coupling is most mated with the chat attribute information of input expression to be detected as far as possible in storehouse, the present embodiment is in advance to pre-
If expression data storehouse in expression carry out Text Flag, obtain text expression, then extract respectively corresponding with text expression
Term vector and term vector corresponding with the chat attribute information of input expression to be detected, so that obtain text expression term vector
Chat attribute term vector, finally by the similarity calculated between the two term vectors, and similarity highest text is expressed one's feelings
The corresponding expression of term vector is used as the error correction expression that most mates with the chat attribute information of input expression to be detected.
The text expression of the present embodiment indication specifically refers to carry out the Text Flag content obtained after Text Flag to expression,
The expression Text Flag that for example one is wailed is " crying ", then Text Flag content of " crying " is exactly text expression.And extract with
During the text corresponding term vector of expression, it is to extract the corresponding term vector of text mark content.Additionally, the present embodiment extract with defeated
Enter expression chat attribute information corresponding term vector when, specifically corresponding with chat attribute information term vector.For example, it is assumed that
Chat attribute information for input expression A includes { s1, s2, s3 }, then corresponding term vector is exactly will to chat in attribute
Each chat property content be converted into term vector, it is assumed that for { c1, c 2, c 3 }, so as to obtain chat attribute term vector
For { c1, c 2, c 3 }.
Alternatively, chat attribute information includes:
One kind or many in chatting object's relation, chatting time, chat theme, chat context and chat geographical location information
Plant combination.
Specifically, the chat attribute information in the present embodiment is not limited to include chatting object's relation, chatting time, chat master
One or more combination in topic, chat context and chat geographical location information, specially self-defined as needed.
Below for one simplify embodiment to the present invention to chat process in input expression carry out automatic error-correcting
Method is illustrated further.
With reference to Fig. 2, in the present embodiment to chat process in input expression carry out the method for automatic error-correcting and include:
Step S201, gathers the chat attribute information of training sample, and wherein, training sample includes that input expression is wrong and is input into
Express one's feelings faultless training sample.
Specifically, substantial amounts of input expression is wrong and the faultless sample of input expression is used as by gathering for the present embodiment
Training sample, and in order to improve the error detection rate for carrying out error detection using error detection model to input expression as far as possible, the present embodiment is chosen
Input expression is wrong and the number of the input faultless training sample of expression as far as possible quite.
Step S202, based on the chat attribute information of training sample and training sample, extracts the characteristic vector of training sample.
Specifically, the present embodiment needs the chat attribute letter for first obtaining training sample before the characteristic vector for extracting training sample
Breath, and when the chat attribute information of training sample is obtained, the present embodiment sets attributes entries first(For example chatting object is closed
System, chatting time, chat theme, chat context and chat geographical position entry etc.), it is then based on the attribute bar for setting
Mesh obtains the chat attribute information of training sample.Assume that the present embodiment attributes entries set in advance include chat theme, chat
Time, chatting object's relation.And the characteristic vector of training sample is extracted after the chat attribute information for getting training sample.Tool
Body ground, this feature vector are made up of the corresponding term vector of chat attribute information.
Step S203, trains grader according to characteristic vector, obtains error detection model.
Specifically, the grader in the present embodiment can be Bayes classifier, and support vector machine classifier, maximum entropy are divided
Class device etc..
Whether step S204, the current input of detection include emoticon, if so, then using emoticon as to be detected defeated
Enter expression.
Specifically, whether the present embodiment real-time detection currently input include emoticon, it is assumed that user is currently input into
Content is " emoticon of happy holiday+Mother's Day blessing ", then when system detectio to the user input " expression of Mother's Day blessing
During symbol ", the emoticon of " emoticon of Mother's Day blessing " is expressed one's feelings as input to be detected.
Step S205, gathers the chat attribute information of input expression to be detected.
As the present embodiment is when the chat attribute information of training sample is gathered, attributes entries set in advance include chatting
Theme, chatting time, chatting object's relation, therefore the chat attribute information of input expression to be detected is being gathered, it is also adopted by above-mentioned
Attributes entries.Assume according to above-mentioned attributes entries, the chat attribute information for collecting input expression to be detected is:{ chat master
The blessing of topic=red-letter day, chatting time=March 8, chatting object's relation=friend }.
Step S206, based on input expression to be detected and the chat attribute information of input expression to be detected, extraction is treated
The characteristic vector of the input expression of detection.
Specifically, the characteristic vector namely input expression to be detected of the input expression to be detected that the present embodiment is extracted and
The combination of the corresponding term vector of attribute information of chatting of input expression to be detected.
Step S207, by the characteristic vector input error detection model of input expression to be detected, judges input table to be detected
Whether feelings are accurate.
Specifically, it is assumed that after the present embodiment is by the characteristic vector input error detection model of input expression to be detected, detect
Input expression is inaccurate, then execution step S208.
Step S208, carries out Text Flag in advance to the expression in default expression data storehouse, obtains text expression.
Specifically, the present embodiment carries out Text Flag to the expression in default expression data storehouse first, so as to obtain use
The text expression of Text Flag.
Step S209, extracts and the text corresponding term vector of expression, obtains text expression term vector.
Specifically, by extracting and each text corresponding term vector of expression respectively, so as to obtain and each expression
Corresponding text expression term vector.
Step S210, extracts term vector corresponding with the chat attribute information of input expression to be detected, obtains chat category
Property term vector.
Specifically, the present embodiment chat attribute term vector to be obtained namely the chat attribute with input expression to be detected
The corresponding term vector of information, namely with { chat theme=red-letter day blessing, chatting time=March 8, chatting object's relation=friend }
Corresponding term vector.
Step S211, the similarity between calculating chat attribute term vector and text expression term vector, and by similarity most
The corresponding expression of high text expression term vector is used as the error correction that most mates with the chat attribute information of input expression to be detected
Expression.
Specifically, it is assumed that by Similarity Measure, the present embodiment obtains the chat attribute letter with input expression to be detected
" emoticon of International Working Woman's Day blessing " of the error correction expression that breath most mates.
Step S212, error correction is expressed one's feelings and replaces input expression to be detected.
Specifically, " emoticon of International Working Woman's Day blessing " is replaced by the error correction expression for being obtained according to step S211, the present embodiment
Change " emoticon of Mother's Day blessing ".
The present invention provide to chat process in input expression carry out automatic error-correcting method, by gather training sample
Chat attribute information;Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;Root
Grader is trained according to characteristic vector, error detection model is obtained and according to error detection model, is judged whether input expression to be detected is accurate
Really, if it is not, then carrying out error correction to input expression to be detected, solving causes to be input into due to wrong choosing expression or misunderstanding expression meanings
Express one's feelings inaccurate technical problem, on the one hand by using error detection model to chat process in input expression realize examining automatically
Mistake, on the other hand inaccurate input expression realize automatic error-correcting.
With reference to Fig. 3, the preferred embodiments of the present invention provide to chat process in input expression carry out automatic error-correcting
Device, including:
Harvester 10, for gathering the chat attribute information of training sample, wherein, training sample includes that input expression is wrong
With the faultless training sample of input expression;
Feature-vector extraction means 20, for the chat attribute information based on training sample and training sample, extract training sample
Characteristic vector;
Training devicess 30, for training grader according to characteristic vector, obtain error detection model;
Whether accurately error correction device 40, for according to error detection model, judging input expression to be detected, if it is not, then to be detected
Input expression carry out error correction.
Alternatively, error correction device 40 includes:
Detection means, for detecting current input whether comprising emoticon, if so, then using emoticon as to be detected defeated
Enter expression;
Chat attribute information harvester, for gathering the chat attribute information of input expression to be detected;
Input expressive features vector extraction device, for the chat based on input expression to be detected and input expression to be detected
Attribute information, extracts the characteristic vector of input expression to be detected;
Judgment means, for the characteristic vector input error detection model by input expression to be detected, judge input table to be detected
Whether feelings are accurate.
Alternatively, error correction device 40 also includes:
Coalignment, in default expression data storehouse, mating the chat attribute information that expresses one's feelings with input to be detected most
The error correction expression of coupling;
Alternative, replaces input expression to be detected for error correction is expressed one's feelings.
Alternatively, coalignment includes:
Identity device, for carrying out Text Flag in advance to the expression in default expression data storehouse, obtains text expression;
Text expression term vector acquisition device, for extraction and the text corresponding term vector of expression, obtains text expression term vector;
Chat attribute term vector acquisition device, for extract word corresponding with the chat attribute information of input expression to be detected to
Amount, obtains chat attribute term vector;
Similarity Measure device, for calculating the similarity between chat attribute term vector and text expression term vector, and by phase
Seemingly the corresponding expression of degree highest text expression term vector is most mated as the chat attribute information that is expressed one's feelings with input to be detected
Error correction expression.
The present invention provide to chat process in input expression carry out the device of automatic error-correcting, by gathering training sample
Chat attribute information;Based on the chat attribute information of training sample and training sample, the characteristic vector of training sample is extracted;Root
Grader is trained according to characteristic vector, error detection model is obtained and according to error detection model, is judged whether input expression to be detected is accurate
Really, if it is not, then carrying out error correction to input expression to be detected, solving causes to be input into due to wrong choosing expression or misunderstanding expression meanings
Express one's feelings inaccurate technical problem, on the one hand by using error detection model to chat process in input expression realize examining automatically
Mistake, on the other hand inaccurate input expression realize automatic error-correcting.
The present embodiment to chat process in input expression carry out automatic error-correcting device specific work process and work
Principle can refer to the present embodiment to chat process in input expression carry out automatic error-correcting method the course of work and work
Principle.
The preferred embodiments of the present invention are these are only, the present invention is not limited to, for those skilled in the art
For member, the present invention can have various modifications and variations.All any modifications that within the spirit and principles in the present invention, is made,
Equivalent, improvement etc., should be included within the scope of the present invention.
Claims (9)
1. a kind of to chat process in input expression carry out automatic error-correcting method, it is characterised in that include:
The chat attribute information of collection training sample, wherein, the training sample includes that input expression is wrong and is input into expression
Faultless training sample;
Based on the chat attribute information of the training sample and the training sample, the characteristic vector of the training sample is extracted;
Grader is trained according to the characteristic vector, error detection model is obtained;
According to the error detection model, whether accurately input expression to be detected is judged, if it is not, then to the input table to be detected
Feelings carry out error correction.
2. according to claim 1 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
According to the error detection model, judge whether input expression to be detected accurately includes:
Whether the current input of detection includes emoticon, and if so, then the emoticon is expressed one's feelings as input to be detected;
Gather the chat attribute information of the input expression to be detected;
Based on the input expression to be detected and the chat attribute information of the input expression to be detected, extract described to be checked
The characteristic vector of the input expression of survey;
The characteristic vector of the input expression to be detected is input into the error detection model, the input expression to be detected is judged
Whether accurate.
3. according to claim 2 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
Error correction is carried out to the input expression to be detected includes:
In default expression data storehouse, mate the error correction that most mates with the chat attribute information of the input expression to be detected
Expression;
The error correction is expressed one's feelings and replaces the input expression to be detected.
4. according to claim 3 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
In default expression data storehouse, mate the error correction expression that most mates with the chat attribute information of the input expression to be detected
Including:
Carry out Text Flag in advance to the expression in default expression data storehouse, obtain text expression;
Extract and the text corresponding term vector of expression, obtain text expression term vector;
Term vector corresponding with the chat attribute information of the input expression to be detected is extracted, chat attribute term vector is obtained;
The similarity between the chat attribute term vector and text expression term vector is calculated, and similarity highest is literary
The corresponding expression of this expression term vector is used as the error correction table most mated with the chat attribute information of the input expression to be detected
Feelings.
5. according to claim 4 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
The chat attribute information includes:
One kind or many in chatting object's relation, chatting time, chat theme, chat context and chat geographical location information
Plant combination.
6. a kind of to chat process in input expression carry out the device of automatic error-correcting, it is characterised in that
Harvester, for gathering the chat attribute information of training sample, wherein, the training sample includes that input expression is wrong
Miss and the faultless training sample of input expression;
Feature-vector extraction means, for the chat attribute information based on the training sample and the training sample, extract institute
State the characteristic vector of training sample;
Training devicess, for training grader according to the characteristic vector, obtain error detection model;
Whether accurately error correction device, for according to the error detection model, judging input expression to be detected, if it is not, then to described
Input expression to be detected carries out error correction.
7. according to claim 6 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
The error correction device includes:
Detection means, for detecting current input whether comprising emoticon, if so, then using the emoticon as to be detected
Input expression;
Chat attribute information harvester, for gathering the chat attribute information of the input expression to be detected;
Input expressive features vector extraction device, for based on the input expression to be detected and the input table to be detected
The chat attribute information of feelings, extracts the characteristic vector of the input expression to be detected;
Judgment means, for the characteristic vector of the input expression to be detected is input into the error detection model, judge described in treat
Whether the input expression of detection is accurate.
8. according to claim 7 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
The error correction device also includes:
Coalignment, in default expression data storehouse, mating the chat attribute letter with the input expression to be detected
The error correction expression that breath most mates;
Alternative, replaces the input expression to be detected for the error correction is expressed one's feelings.
9. according to claim 8 to chat process in input expression carry out automatic error-correcting method, it is characterised in that
The coalignment includes:
Identity device, for carrying out Text Flag in advance to the expression in default expression data storehouse, obtains text expression;
Text expression term vector acquisition device, for extraction and the text corresponding term vector of expression, obtains text expression word
Vector;
Chat attribute term vector acquisition device, corresponding with the chat attribute information of the input expression to be detected for extracting
Term vector, obtains chat attribute term vector;
Similarity Measure device, similar between the chat attribute term vector and text expression term vector for calculating
Degree, and the corresponding expression of term vector that similarity highest text is expressed one's feelings is used as the chat category with the input expression to be detected
Property information most mate error correction expression.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610943311.0A CN106503744A (en) | 2016-10-26 | 2016-10-26 | Input expression in chat process carries out the method and device of automatic error-correcting |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610943311.0A CN106503744A (en) | 2016-10-26 | 2016-10-26 | Input expression in chat process carries out the method and device of automatic error-correcting |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106503744A true CN106503744A (en) | 2017-03-15 |
Family
ID=58321895
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610943311.0A Pending CN106503744A (en) | 2016-10-26 | 2016-10-26 | Input expression in chat process carries out the method and device of automatic error-correcting |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106503744A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107491493A (en) * | 2017-07-22 | 2017-12-19 | 长沙兔子代跑网络科技有限公司 | A kind of intelligence obtains the method and device for running chat record in generation |
CN110188274A (en) * | 2019-05-30 | 2019-08-30 | 口口相传(北京)网络技术有限公司 | Search for error correction method and device |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103823561A (en) * | 2014-02-27 | 2014-05-28 | 广州华多网络科技有限公司 | Expression input method and device |
CN104053131A (en) * | 2013-03-12 | 2014-09-17 | 华为技术有限公司 | Text communication information processing method and related equipment |
CN104063427A (en) * | 2014-06-06 | 2014-09-24 | 北京搜狗科技发展有限公司 | Expression input method and device based on semantic understanding |
CN104463231A (en) * | 2014-12-31 | 2015-03-25 | 合一网络技术(北京)有限公司 | Error correction method used after facial expression recognition content is labeled |
CN104484058A (en) * | 2014-12-19 | 2015-04-01 | 天脉聚源(北京)科技有限公司 | Instant expression image outputting method and instant expression image outputting device |
CN104598127A (en) * | 2014-12-31 | 2015-05-06 | 广东欧珀移动通信有限公司 | Method and device for inserting emoticon in dialogue interface |
US20160110058A1 (en) * | 2011-12-19 | 2016-04-21 | Machine Zone, Inc. | Systems and methods for identifying and suggesting emoticons |
CN106021599A (en) * | 2016-06-08 | 2016-10-12 | 维沃移动通信有限公司 | Emotion icon recommending method and mobile terminal |
-
2016
- 2016-10-26 CN CN201610943311.0A patent/CN106503744A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160110058A1 (en) * | 2011-12-19 | 2016-04-21 | Machine Zone, Inc. | Systems and methods for identifying and suggesting emoticons |
CN104053131A (en) * | 2013-03-12 | 2014-09-17 | 华为技术有限公司 | Text communication information processing method and related equipment |
CN103823561A (en) * | 2014-02-27 | 2014-05-28 | 广州华多网络科技有限公司 | Expression input method and device |
CN104063427A (en) * | 2014-06-06 | 2014-09-24 | 北京搜狗科技发展有限公司 | Expression input method and device based on semantic understanding |
CN104484058A (en) * | 2014-12-19 | 2015-04-01 | 天脉聚源(北京)科技有限公司 | Instant expression image outputting method and instant expression image outputting device |
CN104463231A (en) * | 2014-12-31 | 2015-03-25 | 合一网络技术(北京)有限公司 | Error correction method used after facial expression recognition content is labeled |
CN104598127A (en) * | 2014-12-31 | 2015-05-06 | 广东欧珀移动通信有限公司 | Method and device for inserting emoticon in dialogue interface |
CN106021599A (en) * | 2016-06-08 | 2016-10-12 | 维沃移动通信有限公司 | Emotion icon recommending method and mobile terminal |
Non-Patent Citations (1)
Title |
---|
陆玉清等: "对聊天过程中的输入表情进行自动纠错的方法", 《中文信息学报》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107491493A (en) * | 2017-07-22 | 2017-12-19 | 长沙兔子代跑网络科技有限公司 | A kind of intelligence obtains the method and device for running chat record in generation |
CN110188274A (en) * | 2019-05-30 | 2019-08-30 | 口口相传(北京)网络技术有限公司 | Search for error correction method and device |
CN110188274B (en) * | 2019-05-30 | 2021-06-08 | 口口相传(北京)网络技术有限公司 | Search error correction method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102722709B (en) | Method and device for identifying garbage pictures | |
CN104717124B (en) | A kind of friend recommendation method, apparatus and server | |
CN107146099B (en) | Marketing method and marketing system | |
CN104080054B (en) | A kind of acquisition methods and device of exception point of interest | |
CN102456022B (en) | Short message management method and system | |
CN104717674A (en) | Number attribute recognition method and device, terminal and server | |
CN104040963A (en) | System and methods for spam detection using frequency spectra of character strings | |
CN103473289A (en) | Device and method for completing communication addresses | |
US11429923B2 (en) | Method and device for determining mail path information | |
CN107918629A (en) | The correlating method and device of a kind of alarm failure | |
CN103678531A (en) | Friend recommendation method and friend recommendation device | |
CN104049874A (en) | Touch chirography generation method and system of touch pen | |
CN103580939A (en) | Method and device for detecting abnormal messages based on account number attributes | |
CN104484390A (en) | Zombie fan detecting method facing microblog | |
CN106649410A (en) | Method and device for obtaining chitchat reply content | |
CN105631641A (en) | Flow processing method and apparatus | |
CN106503744A (en) | Input expression in chat process carries out the method and device of automatic error-correcting | |
CN105472618A (en) | Pseudo base station message identification method and pseudo base station message identification device | |
CN109213858B (en) | Automatic identification method and system for network water army | |
CN104268214B (en) | A kind of user's gender identification method and system based on microblog users relation | |
CN103929499A (en) | Internet of things heterogeneous identification recognition method and system | |
CN104102411B (en) | A kind of method for editing text and text editing apparatus | |
CN104065559A (en) | Email marking method and device | |
CN103595614A (en) | User feedback based junk mail detection method | |
CN103634760A (en) | System and method for personal address book management |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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: 20170315 |