CN108038234A - A kind of question sentence template automatic generation method and device - Google Patents
A kind of question sentence template automatic generation method and device Download PDFInfo
- Publication number
- CN108038234A CN108038234A CN201711436114.0A CN201711436114A CN108038234A CN 108038234 A CN108038234 A CN 108038234A CN 201711436114 A CN201711436114 A CN 201711436114A CN 108038234 A CN108038234 A CN 108038234A
- Authority
- CN
- China
- Prior art keywords
- question sentence
- template
- msub
- sentence template
- question
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 239000000463 material Substances 0.000 claims abstract description 66
- 238000005065 mining Methods 0.000 claims abstract description 20
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000012512 characterization method Methods 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 239000013256 coordination polymer Substances 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 238000002360 preparation method Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 5
- 230000002045 lasting effect Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 10
- 201000010099 disease Diseases 0.000 description 7
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 238000013441 quality evaluation Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000009412 basement excavation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000001072 colon Anatomy 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 208000028782 Hereditary disease Diseases 0.000 description 1
- 206010020850 Hyperthyroidism Diseases 0.000 description 1
- 208000026350 Inborn Genetic disease Diseases 0.000 description 1
- 208000024556 Mendelian disease Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 208000006454 hepatitis Diseases 0.000 description 1
- 231100000283 hepatitis Toxicity 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
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
- G06F16/3332—Query translation
- G06F16/3334—Selection or weighting of terms from queries, including natural language queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Abstract
The invention discloses a kind of question sentence template automatic generation method and device, belong to intelligent answer technical field.The described method includes：Prepare question sentence daily record language material；The daily record language material is segmented and part-of-speech tagging；It is named Entity recognition and replacement；Carry out semantic replacement；Frequent item set mining is carried out, generates question sentence template.This method and device not only increase the efficiency of question sentence template generation, substantial saving in manual resource, and the question sentence template of generation can be assessed, autonomous lasting extension question sentence template library, lift the quality of intelligent Answer System knowledge base.
Description
Technical field
The present invention relates to intelligent answer technical field, more particularly to a kind of question sentence template automatic generation method and device.
Background technology
At present, more and more enterprises can undertake a large number of users after-sale service or pre-sales work of consultation.Due to number of users
Exponential increase, all users consulting is carried out by answer can expend substantial amounts of manual resource completely by the way of artificial, and
Many knowledge points Relatively centralized, manually replys and usually contains a large amount of duplications of labour, so that intelligent Answer System comes into being, intelligence
Question answering system can be answered automatically for problem input by user, and efficiency is increased substantially.
The technical principle of intelligent Answer System has based on modes such as question sentence template matches, knowledge library searchings.Wherein, question sentence mould
Plate matching technique is one of most popular technology, question sentence template refer to by question sentence is identified replace it after formed
Specific symbol label sequence, corresponding answer is increased to question sentence template problem, is being run into and the same or similar degree of template
During the problem of very high, template matching technique can carry out matching answer to the problem.The difficult point of question sentence template matching technique be as
What efficient and sustainable generation question sentence template.The generation of traditional question sentence template for specific clause, it is necessary to manually carry out mould
Plate set, it is not only cumbersome and can spreadability it is poor；When knowledge base update, it is also desirable to manually to that cannot cover in template library
Question sentence template carry out the setting and assessment of new template, maintainable and self-teaching is poor.In presently disclosed related patents,
Do not find also to make correspondingly modified technical solution for above technical problem, such as patent application 201611076382.1 (《One
The method and device of kind automatic question answering template matches》), pass through the template problem set of the corresponding each participle that determines to wait to answer a question
Subset, obtains matching problem to be answered a question, and improves template matches efficiency and the accuracy of automatically request-answering system, but not
It is related to template and automatically generates and generate template quality evaluation problem.
The content of the invention
In order to solve problem of the prior art, an embodiment of the present invention provides a kind of question sentence template automatic generation method and dress
Put.The technical solution is as follows：
First aspect, there is provided a kind of question sentence template automatic generation method, the described method includes：
Prepare question sentence daily record language material；The daily record language material is segmented and part-of-speech tagging；Be named Entity recognition and
Replace；Carry out semantic replacement；Frequent item set mining is carried out, generates question sentence template.
With reference to first aspect, in the first possible implementation, question sentence daily record language material is prepared, including：
Question sentence daily record language material is obtained, and question sentence daily record language material is pre-processed, including punctuation mark removes, illegal symbol
Remove, the conversion of word capital and small letter.
With reference to first aspect, the daily record language material may be segmented and part-of-speech tagging in implementation at second,
Including：
The daily record language material is segmented with reference to the segmenting method of industry dictionary.
With reference to first aspect, in the third possible implementation, Entity recognition and replacement are named, including：
Entity knowledge is named to the general entity including time, numeral and/or place name occurred in question sentence daily record language material
Not, and by the general entity it is substituted for corresponding entity tag.
With reference to first aspect, in the 4th kind of possible implementation, semantic replacement is carried out, including：
Word after question sentence participle in question sentence daily record language material is searched for by semantic net, will be same or similar according to the paraphrase of word
The word of paraphrase, which is abstracted, is unified for label, and is accordingly replaced, and generates by name entity and semantic replaced semantic concept structure
Into symbol label sequence.
The 4th kind of possible implementation with reference to first aspect, in the 5th kind of possible implementation, carries out frequent item set
Excavate, generate question sentence template, including：
By given threshold scope, frequent item set is obtained from the candidate of question sentence language material daily record, generates question sentence template.
The 5th kind of possible implementation with reference to first aspect, in the 6th kind of possible implementation, carries out frequent item set
Excavate, generate question sentence template, including：
According to default frequency threshold range and default item collection length threshold range, using predetermined association rule-based algorithm from described
Frequent item set is screened in symbol label sequence, the sequence formed according to the default sequence of item is to generate question sentence template.
The five to six kind of possible implementation with reference to first aspect, it is described in the seven to eight kind of possible implementation
Method further includes：
Sentence vector characterization is carried out to the question sentence under the question sentence template that filters out using default sentence vector model；
The cluster compactness of the question sentence template is calculated using following calculation formula：
According to default template cluster compactness threshold value, the question sentence mould that cluster compactness is more than the tight ness rating threshold value is filtered out
Plate；
The question sentence template filtered out is subjected to lookup contrast in template library, if the question sentence mould filtered out is not present in template library
Plate, the question sentence template filtered out is preserved to template library；
Wherein, in calculation formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor j-th of question sentence
The sentence vector of i-th of question sentence, W under templatejFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor
For all long summations of vectorial mould of the corresponding cluster of j-th of question sentence template, i, j are the integer more than or equal to 1.
The seven to eight kind of possible implementation with reference to first aspect, it is described in the nine to ten kind of possible implementation
Default sentence vector model is deep learning encoder model Skip-Thoughts.
The seven to eight kind of possible implementation with reference to first aspect, in the 11st to 12 kind of possible implementation,
The method further includes：
Increase answer corresponding with the question sentence template filtered out, complete question sentence template is formed with the question sentence template filtered out
Question and answer pair, preserve to template library.
A kind of second aspect, there is provided question sentence template automatically generating device, it is characterised in that including：
Preparation module, for preparing question sentence daily record language material；Participle and part-of-speech tagging module, for carrying out participle and part of speech mark
Note；Entity recognition module is named, for being named Entity recognition and replacement；Semantic replacement module, for carrying out semantic replacement；
Frequent item set mining module, for carrying out frequent item set mining, generates question sentence template.
With reference to second aspect, in the first possible implementation, the preparation module includes acquisition module and pretreatment
Module, the acquisition module are used to obtain question sentence daily record language material, and the pretreatment module is used to carry out question sentence daily record language material pre-
Processing, including punctuation mark removes, illegal symbol removes, the conversion of word capital and small letter.
With reference to second aspect, in second of possible implementation, the name Entity recognition module is used for：To question sentence day
The general entity including time, numeral and/or place name occurred in will language material is named Entity recognition, and by the general reality
Body is substituted for corresponding entity tag.
With reference to second aspect, in the third possible implementation, the semanteme replacement module is used for：By question sentence daily record language
Word in material after question sentence participle is searched for by semantic net, is abstracted the word of same or similar paraphrase according to the paraphrase of word and is unified for mark
Label, and accordingly replaced, generate the symbol label sequence being made of name entity and semantic replaced semantic concept.
With reference to the third possible implementation of second aspect, in the 4th kind of possible implementation, the frequent item set
Module is excavated to be used for：According to default frequency threshold range and default item collection length threshold range, predetermined association rule-based algorithm is utilized
Frequent item set is screened from the symbol label sequence, the sequence formed according to the default sequence of item is to generate question sentence template.
With reference to the three to four kind of possible implementation of second aspect, in the five to six possible implementation, the dress
Put and further include：
The vectorial characterization module of sentence, for using default sentence vector model to the question sentence progress sentence under the question sentence template that filters out
Vector characterization；
Cluster compactness computing module, the cluster for calculating the question sentence template using following calculation formula are close
Degree：
Screening module, for according to default template cluster compactness threshold value, it is close more than this to filter out cluster compactness
Spend the question sentence template of threshold value；
Determine preserving module, the question sentence template for that will filter out carries out lookup contrast in template library, if template library is not
In the presence of the question sentence template filtered out, the question sentence template filtered out is preserved to template library；
Wherein, in calculation formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor j-th of question sentence
The sentence vector of i-th of question sentence, W under templatejFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor
For all long summations of vectorial mould of the corresponding cluster of j-th of question sentence template, i, j are the integer more than or equal to 1.
It is described in the seven to eight kind of possible implementation with reference to the five to six kind of possible implementation of second aspect
Default sentence vector model is deep learning encoder model Skip-Thoughts.
It is described in the nine to ten kind of possible implementation with reference to the five to six kind of possible implementation of second aspect
Device further includes：
Answer add module, the corresponding answer of question sentence template for increasing with filtering out, with the question sentence template filtered out
Complete question sentence template question and answer pair are formed, are preserved to template library.
The beneficial effect that technical solution provided in an embodiment of the present invention is brought is：
1st, by semantic replacement step, the word of more justice of word one is carried out by abstract unification according to paraphrase, so as to add semanteme
Generalization ability；
2nd, by carrying out frequent item set mining, frequent item set is looked for from candidate, question sentence template is generated, improves question sentence mould
The efficiency of plate generation, substantial saving in manual resource；
3rd, by according to default item collection frequency threshold range and default item collection length threshold range, being screened from frequent item set
Go out satisfactory item collection, to generate question sentence template, the sentence with similar structure and public word sequence can be clustered out, obtain
The question sentence template of better quality；
4th, default sentence vector model sentence vector characterization, calculating cluster compactness and the satisfactory question sentence mould of screening are utilized
Plate, can realize semantic dimension to generate template quality evaluation, so as to obtain the high quality question sentence template of accuracy higher；
5th, the question sentence template filtered out is subjected to lookup contrast in template library, if the question sentence filtered out is not present in template library
Template, the question sentence template filtered out is preserved to template library, easy to generate effective renewal of the template in template library；
6th, increase answer corresponding with the question sentence template filtered out, complete question sentence mould is formed with the question sentence template filtered out
Plate question and answer pair, preserve to template library, it is ensured that the integrality of template question and answer pair in template library, realization automatically generate answering for question sentence template
Case matches.
In general, question sentence template automatic generation method provided in an embodiment of the present invention and device, can not only be efficiently
Question sentence template is automatically generated, and quality evaluation, autonomous lasting extension or renewal question sentence can be carried out to the question sentence template of generation
Template library, lifts the quality of intelligent Answer System knowledge base, in intelligent answer etc. the technical field of customer service can be needed to carry out extensive
Popularization and application.
Brief description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for
For those of ordinary skill in the art, without creative efforts, other can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is the question sentence template automatic generation method flow chart that the embodiment of the present invention 1 provides；
Fig. 2 is the question sentence template automatic generation method flow chart that the embodiment of the present invention 2 provides；
Fig. 3 is the question sentence template automatically generating device structure diagram that the embodiment of the present invention 3 provides.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached in the embodiment of the present invention
Figure, is clearly and completely described the technical solution in the embodiment of the present invention, it is clear that described embodiment is only this
Invention part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art exist
All other embodiments obtained under the premise of creative work are not made, belong to the scope of protection of the invention.
It should be noted that term " first ", " second " are only used for description purpose, and it is not intended that instruction or hint phase
To importance or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can be with
Express or implicitly include one or more this feature.In the description of the present invention, " multiple " be meant that two with
On, unless otherwise specifically defined.
The purpose of the present invention is to solve in the template matching technique of intelligent Answer System, the generation of question sentence template and
The problem of assessment.From the question sentence daily record of question answering system, question sentence template automatic generation method and dress provided in an embodiment of the present invention
Putting by the Mining Frequent Itemsets Based from question sentence daily record language material, can automatically generate the question sentence template candidate collection of question answering system, so that real
Now automatically generate question sentence template and template quality is assessed, establish the self-study mechanism of intelligent Answer System.With tradition
The method that template is established using artificial rule is compared, and this method and device not only increase the efficiency of question sentence template generation, largely
Manual resource has been saved, and the question sentence template of generation can have been assessed, autonomous lasting extension question sentence template library, lifts intelligence
The quality of energy question answering system knowledge base, therefore question sentence template automatic generation method provided in an embodiment of the present invention and device can be in intelligence
Energy question and answer etc. need the technical field of customer service widely promote and apply.
Embodiment 1
Fig. 1 is question sentence template automatic generation method flow chart provided in an embodiment of the present invention.As shown in Figure 1, the present invention is real
The question sentence template automatic generation method of example offer is provided, is comprised the following steps：
101st, question sentence daily record language material is prepared.
Specifically, obtaining question sentence daily record language material, and question sentence daily record language material is pre-processed, include but not limited to punctuate symbol
Number remove, illegal symbol remove, word capital and small letter conversion etc..
102nd, daily record language material is segmented and part-of-speech tagging.
Specifically, the segmenting method with reference to industry dictionary segments daily record language material.As needed or industry difference,
Different types of industry dictionary can be created, during particularly with regard to specific vertical industry, using the participle side for combining industry dictionary
Method segments corresponding daily record language material, can obtain preferable participle effect.
103rd, Entity recognition and replacement are named.
Specifically, the general entity including time, numeral and/or place name occurred in question sentence daily record language material is named
Entity recognition, and general entity is substituted for corresponding entity tag.
It should be noted that the above-mentioned pretreatment of daily record language material, participle and part-of-speech tagging and name Entity recognition and replacement
It is the more commonly used Sentence analysis treatment technology of natural language processing field, any possible skill in the prior art can also be used
Art means or mode realize above-mentioned several processes, to avoid burden, do not describe in detail herein.
104th, semantic replacement is carried out.
Specifically, the word after question sentence participle in question sentence daily record language material is searched for by semantic net, according to the paraphrase of word by phase
The word of same or similar paraphrase, which is abstracted, is unified for label, and is accordingly replaced, and generates by name entity and semantic replaced language
The symbol label sequence that adopted concept is formed.Here semantic net is preferably by Chinese semantic net HowNet.The mesh of the operating process
, by the generalization ability that semantic abstraction increase is semantic, cannot be found in sentence in semantic net or cannot be real by naming
The word of body identification can directly be ignored.
105th, frequent item set mining is carried out, generates question sentence template.
Specifically, by given threshold scope, frequent item set is obtained from the candidate of question sentence language material daily record, generation is asked
Sentence template.Template excavation is carried out to the daily record language material for being converted into symbol sebolic addressing, the key of template generation is asked from largely different
Frequent item set is clustered out in label symbol sebolic addressing after sentence conversion, these frequent item sets can express clause to a certain extent
Diaphyseal portion, so as to realize automatically generating for question sentence template according to the frequent item set clustered out.Preferably, according to default frequency threshold
It is worth scope and default item collection length threshold range, desired frequency is obtained from symbol label sequence using predetermined association rule-based algorithm
Numerous item collection, to generate question sentence template.That is, frequent item set mining is carried out using predetermined association rule-based algorithm, from candidate item
Collection looks for frequent item set.Here, predetermined association rule-based algorithm can be selected using any possible phase in the prior art as needed
Association rule algorithm is closed, preferably by Apriori algorithm.
Embodiment 2
Fig. 2 is the question sentence template automatic generation method flow chart that the embodiment of the present invention 2 provides.As shown in Fig. 2, the present invention is real
The question sentence template automatic generation method of example offer is provided, is comprised the following steps：
201st, question sentence daily record language material is obtained, and question sentence daily record language material is pre-processed, includes but not limited to punctuation mark
Remove, illegal symbol removes, the conversion of word capital and small letter.
It should be noted that the acquisition methods of the language material of question sentence daily record here can use any in the prior art possible obtain
Method is taken, the embodiment of the present invention is not particularly limited to it；The process or mode for realizing the pretreatment of daily record language material are also not necessarily limited to
The above, can use any possible technological means or mode in the prior art, not repeat one by one herein.
202nd, the segmenting method with reference to industry dictionary segments the daily record language material and part-of-speech tagging.
As needed or industry difference, can create different types of industry dictionary, particularly with regard to specific vertical row
During industry, corresponding daily record language material is segmented using the segmenting method for combining industry dictionary, preferable participle effect can be obtained.
It should be noted that above-mentioned 202 step is not limited to aforesaid operations mode, any possibility in the prior art can be used
Mode or method, the embodiment of the present invention is not particularly limited to it.
203rd, reality is named to the general entity including time, numeral and/or place name occurred in question sentence daily record language material
Body identifies, and general entity is substituted for corresponding entity tag.
It should be noted that above-mentioned name Entity recognition and replacement can also use any possible technology in the prior art
Means or mode are realized, to avoid burden, are not described in detail herein.
204th, the word after question sentence participle in question sentence daily record language material is searched for by semantic net, according to the paraphrase of word by identical or
The word of similar paraphrase, which is abstracted, is unified for label, and is accordingly replaced, and generates general by name entity and semantic replaced semanteme
Read the symbol label sequence formed.
Due to there are the more words of a substantial amounts of justice in Chinese, it is desirable to specific word is abstracted into a representation of word meaning,
To increase the generalization ability of semanteme.Here Chinese semantic net HowNet is preferably used and carries out semantic abstraction replacement.Specific behaviour
It is to be searched the word after participle in semantic net as mode, there is the paraphrase to word in semantic net, will be identical according to the paraphrase of word
Or the word of similar paraphrase is abstracted and is unified for label, and accordingly replaced.For example, definition of the word " hepatitis " in HowNet is
" disease ", and it is also " disease " that " flu " is corresponding in semantic net, the word system that these represent " disease " in sentence
One is substituted for " disease ", so as to achieve the purpose that semantic abstraction.And for that cannot be found in sentence in semantic net or cannot
By naming the word of Entity recognition, can directly ignore.After the step process, language material question sentence changes order already issued from contamination sequence
Name entity and semantic replaced symbol sebolic addressing.
205th, according to default frequency threshold range and default item collection length threshold range, using predetermined association rule-based algorithm from
Frequent item set is screened in the symbol label sequence, the sequence formed according to the default sequence of item is to generate question sentence template.
Specifically, carrying out template excavation to the daily record language material for being converted into symbol sebolic addressing, the key of template generation is from a large amount of
Frequent item set is clustered out in label symbol sebolic addressing after different question sentence conversion, these frequent item sets can to a certain extent can be
Clause diaphyseal portion is expressed, so as to realize automatically generating for question sentence template according to the frequent item set clustered out.That is, using
Predetermined association rule-based algorithm carries out frequent item set mining, and frequent item set is looked for from candidate.Here, predetermined association rule-based algorithm can
With selection as needed using any possible related association rule algorithm in the prior art, preferably by Apriori algorithm.
In order to obtain relatively good generalization ability and semantic tight ness rating, can to the template quality that step before generates into
Row assessment, so as to obtain the question sentence template of better quality.According to predetermined sequence frequency threshold range and predetermined sequence length threshold
Scope, selects satisfactory corresponding candidate template sequence.Specifically, two threshold value indexs at this, first, the sequence is not
With the frequency k1 occurred in language material question sentence, second, sequence length k2, second, the frequency that the sequence occurs in different language material question sentences
Secondary k2.K1 and k2 generally can empirically be set, wherein preferably, k1 is arranged between [3,5], the reason is that length is less than 3
Although template generalization ability it is stronger, semantic tight ness rating is relatively low；Length is preferable more than 5 template semanteme tight ness rating, still
Generalization ability is poor.In the case of appropriate threshold range, this step can be clustered out with similar structure and public word sequence
Sentence, obtain the question sentence template of better quality.
For example, there is so several question sentences in daily record language material, " may I ask hereditary disease can buy", " have a hyperthyroidism can
To buy", " there is can insuring for mild diabetes”.By before the step of handle, three sentences are converted into following respectively
Three symbol sebolic addressings：
[question verb_you disease question_feasible apply_v question_polar]；
[disease question_feasible apply_v question_polar]；
[disease question_feasible apply_v]。
For convenience of description, by the semantic concept in three sequences, with letter, " a b c d e " etc. are replaced respectively.Sequence is changed
For [a b c d e f], [c d e f], [c d e], passes through Frequent Itemsets Mining Algorithm, that is, predetermined association rule-based algorithm, difference
The threshold value of the frequency and item collection length is set, the threshold value for setting the frequency and item collection here is all 3, is calculated in sequence first
Existing frequent item set is [c d e:3], [c d:3], [c e:3], [c:3], [d:3], [e:3] etc., only list total herein
Sequence in occur the combined sequence of 3 times altogether.What the alphabetical sequence combination in sequence before colon represented is asked in difference
Frequent item set in sentence sequence, digitized representation behind colon is the frequency that the frequent item set occurs in language material totality.Example
Such as sequence [c d e], in three sequences of citing, all sequentially occurred with this.[c d e] this sequence can so be obtained
Meet the requirements, the template that this symbol sebolic addressing can be ultimately produced as three words of this citing.
It should be noted that above-mentioned 205 step carries out frequent item set mining using predetermined association rule-based algorithm, from candidate item
Collection looks for frequent item set, and the illustration of generation question sentence template is only exemplary, without departing from the step of the embodiment of the present invention
In the case of specific inventive concept, it can be subject to using other any possible processes or mode, the embodiment of the present invention
It is particularly limited to.
And in order to further improve the quality of question sentence template, it can tied by the template taken out in above-mentioned steps sentence
Characterize similar question sentence in structure feature, but under template corresponding difference question sentences semantically differ surely fully meet it is similar
Property, it is also necessary to further assessed in semantic dimension, implementation step 206~209 as described below of the process.
206th, the question sentence under the question sentence template that filters out is carried out using deep learning encoder model Skip-Thoughts
Sentence vector characterization.
The step carries out cluster compactness to the question sentence of cluster out and calculates, it is ensured that the different question sentences under a template have
The Semantic Similarity of height, so as to be assessed on semantic dimension the template of generation.The step uses the mode of sentence vector
Question sentence is characterized, sentence vector model uses the Skip-Thoughts algorithms that Google increases income, which is non-supervisory model, will
Sentence expression can represent semantic well into the vector of a fixed dimension under large-scale corpus.The model is off-line training,
In training process based on usage log language material term vector, when unregistered word is run into, outside can be preferably combined
Chinese wikipedia language material as word extend.
207th, the cluster compactness of question sentence template is calculated using following calculation formula：
The cluster compactness of the different question sentence templates of multiple classifications can be calculated by the calculation formula.Wherein, calculate
In formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor i-th question sentence under j-th of question sentence template
Sentence vector, WjFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor for j-th of question sentence template pair
The long summation of the vectorial mould of all of the cluster answered, i, j are the integer more than or equal to 1.
208th, according to default template cluster compactness threshold value, cluster compactness asking more than the tight ness rating threshold value is filtered out
Sentence template.Exemplarily, the cluster compactness threshold value k3 of definition template, as step is produced candidate template assessment according to
According to, filter out cluster compactness be more than threshold value template, by the process can realize semantic dimension to generate template matter
Amount assessment, so as to obtain the high quality question sentence template of accuracy higher.Preferably, the initial value setting of threshold value k3 here, can be with
To randomly selecting segment template in original template library, each corresponding cluster compactness is calculated, is then averaged.
It should be noted that above-mentioned 208 step is not limited to aforesaid operations mode, any possibility in the prior art can be used
Mode or method, the embodiment of the present invention is not particularly limited to it.
209th, the question sentence template filtered out is subjected to lookup contrast in template library, if template library be not present filter out to ask
Sentence template, the question sentence template filtered out is preserved to template library.
210th, increase answer corresponding with the question sentence template filtered out, complete question sentence is formed with the question sentence template filtered out
Template question and answer pair, preserve to template library.
It should be noted that during above-mentioned 209~210 step operation, without departing from the step of the embodiment of the present invention
, can be using any possible process or mode, the embodiment of the present invention be not right in the prior art in the case of specific inventive concept
It is particularly limited to.
Embodiment 3
Fig. 3 is the question sentence template automatically generating device structure diagram that the embodiment of the present invention 3 provides.As shown in figure 3, this hair
The question sentence template automatically generating device that bright embodiment provides, including：
Preparation module 1, for preparing question sentence daily record language material.Specifically, preparation module 1 includes acquisition module and pretreatment mould
Block, the acquisition module are used to obtain question sentence daily record language material, and the pretreatment module is used to locate question sentence daily record language material in advance
Reason, including label symbol removes, illegal symbol removes, the conversion of word capital and small letter.
Participle and part-of-speech tagging module 2, for being segmented and part-of-speech tagging.Specifically, as needed or industry is not
Together, different types of industry dictionary can be created, during particularly with regard to specific vertical industry, participle and part-of-speech tagging module 2 are adopted
Corresponding daily record language material is segmented with the segmenting method for combining industry dictionary, preferable participle effect can be obtained.
Entity recognition module 3 is named, for being named Entity recognition and replacement.Specifically, Entity recognition module 3 is named
For：Entity recognition is named to the general entity including time, numeral and/or place name occurred in question sentence daily record language material,
And the general entity is substituted for corresponding entity tag.
Semantic replacement module 4, for carrying out semantic replacement.Semantic replacement module 4 is used for：By question sentence in question sentence daily record language material
Word after participle is searched for by semantic net, is abstracted the word of same or similar paraphrase according to the paraphrase of word and is unified for label, goes forward side by side
Row is corresponding to be replaced, and generates the symbol label sequence being made of name entity and semantic replaced semantic concept.
Frequent item set mining module 5, for carrying out frequent item set mining, generates question sentence template.Specifically, frequent item set is dug
Pick module 5 is used for：According to default frequency threshold range and default item collection length threshold range, using predetermined association rule-based algorithm from
Frequent item set is screened in symbol label sequence, the sequence formed according to the default sequence of item is to generate question sentence template.
Preferably, above-mentioned question sentence template automatically generating device further includes：
The vectorial characterization module 6 of sentence, for using default sentence vector model to the question sentence progress under the question sentence template that clusters out
Sentence vector characterization；Preferably, it is deep learning encoder model Skip-Thoughts to preset sentence vector model.
Cluster compactness computing module 7, the cluster for calculating the question sentence template using following calculation formula are close
Degree：
Wherein, in calculation formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor j-th of question sentence
The sentence vector of i-th of question sentence, W under templatejFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor
For all long summations of vectorial mould of the corresponding cluster of j-th of question sentence template, i, j are the integer more than or equal to 1.
Screening module 8, for according to default template cluster compactness threshold value, it is close more than this to filter out cluster compactness
Spend the question sentence template of threshold value；
Determine preserving module 9, the question sentence template for that will filter out carries out lookup contrast in template library, if template library is not
In the presence of the question sentence template filtered out, the question sentence template filtered out is preserved to template library；
Additionally preferably, above-mentioned question sentence template automatically generating device further includes：
Answer add module 10, the corresponding answer of question sentence template for increasing with filtering out, with the question sentence mould filtered out
Plate forms complete question sentence template question and answer pair, preserves to template library.
It should be noted that：The question sentence template automatically generating device that above-described embodiment provides is given birth to automatically in progress question sentence template
, can as needed will be above-mentioned only with the division progress of above-mentioned each function module for example, in practical application during into business
Function distribution is completed by different function module, i.e., the internal structure of device is divided into different function modules, with complete with
The all or part of function of upper description.In addition, question sentence template automatically generating device and question sentence template that above-described embodiment provides
Automatic generation method embodiment belongs to same design, its specific implementation process refers to embodiment of the method, and which is not described herein again.
In conclusion question sentence template automatic generation method provided in an embodiment of the present invention and device, relative to the prior art
With following beneficial aspects：
1st, by semantic replacement step, the word of more justice of word one is carried out by abstract unification according to paraphrase, so as to add semanteme
Generalization ability；
2nd, by carrying out frequent item set mining, frequent item set is looked for from candidate, question sentence template is generated, improves question sentence mould
The efficiency of plate generation, substantial saving in manual resource；
3rd, by according to predetermined sequence frequency threshold range and predetermined sequence length threshold scope, being screened from frequent item set
Go out satisfactory item collection, to generate question sentence template, the sentence with similar structure and public word sequence can be clustered out, obtain
The question sentence template of better quality；
4th, default sentence vector model sentence vector characterization, calculating cluster compactness and the satisfactory question sentence mould of screening are utilized
Plate, can realize semantic dimension to generate template quality evaluation, so as to obtain the high quality question sentence template of accuracy higher；
5th, the question sentence template filtered out is subjected to lookup contrast in template library, if the question sentence filtered out is not present in template library
Template, the question sentence template filtered out is preserved to template library, easy to generate effective renewal of the template in template library；
6th, increase answer corresponding with the question sentence template filtered out, complete question sentence mould is formed with the question sentence template filtered out
Plate question and answer pair, preserve to template library, it is ensured that the integrality of template question and answer pair in template library, realization automatically generate answering for question sentence template
Case matches.
In general, question sentence template automatic generation method provided in an embodiment of the present invention and device, can not only be efficiently
Question sentence template is automatically generated, and quality evaluation, autonomous lasting extension or renewal question sentence can be carried out to the question sentence template of generation
Template library, lifts the quality of intelligent Answer System knowledge base, in intelligent answer etc. the technical field of customer service can be needed to carry out extensive
Popularization and application.
One of ordinary skill in the art will appreciate that hardware can be passed through by realizing all or part of step of above-described embodiment
To complete, relevant hardware can also be instructed to complete by program, the program can be stored in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only storage, disk or CD etc..
Above-mentioned all optional technical solutions, can use any combination to form the alternative embodiment of the present invention, herein no longer
Repeat one by one.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on, should all be included in the protection scope of the present invention.
Claims (18)
- A kind of 1. question sentence template automatic generation method, it is characterised in that the described method includes：Prepare question sentence daily record language material；The daily record language material is segmented and part-of-speech tagging；It is named Entity recognition and replacement；Carry out semantic replacement；Frequent item set mining is carried out, generates question sentence template.
- 2. according to the method described in claim 1, it is characterized in that, prepare question sentence daily record language material, including：Question sentence daily record language material is obtained, and question sentence daily record language material is pre-processed, including punctuation mark removes, illegal symbol is gone Remove, the conversion of word capital and small letter.
- 3. according to the method described in claim 1, it is characterized in that, the daily record language material is segmented and part-of-speech tagging, bag Include：The daily record language material is segmented with reference to the segmenting method of industry dictionary.
- 4. according to the method described in claim 1, it is characterized in that, be named Entity recognition and replacement, including：Entity recognition is named to the general entity including time, numeral and/or place name occurred in question sentence daily record language material, and The general entity is substituted for corresponding entity tag.
- 5. according to the method described in claim 1, it is characterized in that, carry out semantic replacement, including：Word after question sentence participle in question sentence daily record language material is searched for by semantic net, according to the paraphrase of word by same or similar paraphrase Word be abstracted and be unified for label, and accordingly replaced, generate what is be made of name entity and semantic replaced semantic concept Symbol label sequence.
- 6. according to the method described in claim 5, it is characterized in that, carry out frequent item set mining, generate question sentence template, including：By given threshold scope, frequent item set is obtained from the candidate of question sentence language material daily record, generates question sentence template.
- 7. according to the method described in claim 6, it is characterized in that, carry out frequent item set mining, generate question sentence template, including：According to default frequency threshold range and default item collection length threshold range, using predetermined association rule-based algorithm from the symbol Frequent item set is screened in sequence label, the sequence formed according to the default sequence of item is to generate question sentence template.
- 8. the method according to claim 6 or 7, it is characterised in that the method further includes：Sentence vector characterization is carried out using the question sentence of question sentence template of the default sentence vector model to filtering out；The cluster compactness of the question sentence template is calculated using following calculation formula：<mrow> <msub> <mover> <mrow> <mi>C</mi> <mi>P</mi> </mrow> <mo>&OverBar;</mo> </mover> <mi>j</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mo>|</mo> <msub> <mi>&Omega;</mi> <mi>j</mi> </msub> <mo>|</mo> </mrow> </mfrac> <munder> <mo>&Sigma;</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>&Element;</mo> <msub> <mi>&Omega;</mi> <mi>j</mi> </msub> </mrow> </munder> <mo>|</mo> <mo>|</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>W</mi> <mi>j</mi> </msub> <mo>|</mo> <mo>|</mo> </mrow>According to default template cluster compactness threshold value, the question sentence template that cluster compactness is more than the tight ness rating threshold value is filtered out；The question sentence template filtered out is subjected to lookup contrast in template library, if the question sentence template filtered out is not present in template library, The question sentence template filtered out is preserved to template library；Wherein, in calculation formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor j-th of question sentence template The sentence vector of lower i-th of question sentence, WjFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor for jth All long summations of vectorial mould of the corresponding cluster of a question sentence template, i, j are the integer more than or equal to 1.
- 9. according to the method described in claim 8, it is characterized in that, the default sentence vector model is deep learning encoder mould Type Skip-Thoughts.
- 10. according to the method described in claim 8, it is characterized in that, the method further includes：Increase answer corresponding with the question sentence template filtered out, complete question sentence template question and answer are formed with the question sentence template filtered out It is right, preserve to template library.
- A kind of 11. question sentence template automatically generating device, it is characterised in that including：Preparation module, for preparing question sentence daily record language material；Participle and part-of-speech tagging module, for being segmented and part-of-speech tagging；Entity recognition module is named, for being named Entity recognition and replacement；Semantic replacement module, for carrying out semantic replacement；Frequent item set mining module, for carrying out frequent item set mining, generates question sentence template.
- 12. according to the devices described in claim 11, it is characterised in that the preparation module includes acquisition module and pretreatment mould Block, the acquisition module are used to obtain question sentence daily record language material, and the pretreatment module is used to locate question sentence daily record language material in advance Reason, including punctuation mark removes, illegal symbol removes, the conversion of word capital and small letter.
- 13. according to the devices described in claim 11, it is characterised in that the name Entity recognition module is used for：To question sentence day The general entity including time, numeral and/or place name occurred in will language material is named Entity recognition, and by the general reality Body is substituted for corresponding entity tag.
- 14. according to the devices described in claim 11, it is characterised in that the semanteme replacement module is used for：By question sentence daily record language Word in material after question sentence participle is searched for by semantic net, is abstracted the word of same or similar paraphrase according to the paraphrase of word and is unified for mark Label, and accordingly replaced, generate the symbol label sequence being made of name entity and semantic replaced semantic concept.
- 15. device according to claim 14, it is characterised in that the frequent item set mining module is used for：According to default Frequency threshold range and default item collection length threshold range, are sieved using predetermined association rule-based algorithm from the symbol label sequence Frequent item set is selected, the sequence formed according to the default sequence of item is to generate question sentence template.
- 16. the device according to claims 14 or 15, it is characterised in that described device further includes：The vectorial characterization module of sentence, for carrying out sentence vector table using the question sentence of question sentence template of the default sentence vector model to filtering out Sign；Cluster compactness computing module, for calculating the cluster compactness of the question sentence template using following calculation formula：<mrow> <msub> <mover> <mrow> <mi>C</mi> <mi>P</mi> </mrow> <mo>&OverBar;</mo> </mover> <mi>j</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mo>|</mo> <msub> <mi>&Omega;</mi> <mi>j</mi> </msub> <mo>|</mo> </mrow> </mfrac> <munder> <mo>&Sigma;</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>&Element;</mo> <msub> <mi>&Omega;</mi> <mi>j</mi> </msub> </mrow> </munder> <mo>|</mo> <mo>|</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>W</mi> <mi>j</mi> </msub> <mo>|</mo> <mo>|</mo> </mrow>Screening module, for according to default template cluster compactness threshold value, filtering out cluster compactness and being more than the tight ness rating threshold The question sentence template of value；Determine preserving module, the question sentence template for that will filter out carries out lookup contrast in template library, if template library is not present The question sentence template filtered out, the question sentence template filtered out is preserved to template library；Wherein, in calculation formula, CPjCluster compactness for j-th of question sentence template being calculated, XiFor j-th of question sentence template The sentence vector of lower i-th of question sentence, WjFor all vectorial average values of the corresponding cluster of j-th of question sentence template；ΩjFor for jth All long summations of vectorial mould of the corresponding cluster of a question sentence template, i, j are the integer more than or equal to 1.
- 17. device according to claim 16, it is characterised in that the default sentence vector model is deep learning encoder Model Skip-Thoughts.
- 18. device according to claim 16, it is characterised in that described device further includes：Answer add module, the corresponding answer of question sentence template for increasing with filtering out, forms with the question sentence template filtered out Complete question sentence template question and answer pair, preserve to template library.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711436114.0A CN108038234B (en) | 2017-12-26 | 2017-12-26 | Automatic question template generating method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711436114.0A CN108038234B (en) | 2017-12-26 | 2017-12-26 | Automatic question template generating method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108038234A true CN108038234A (en) | 2018-05-15 |
| CN108038234B CN108038234B (en) | 2021-06-15 |
Family
ID=62101304
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711436114.0A Active CN108038234B (en) | 2017-12-26 | 2017-12-26 | Automatic question template generating method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108038234B (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108776677A (en) * | 2018-05-28 | 2018-11-09 | 深圳前海微众银行股份有限公司 | Creation method, equipment and the computer readable storage medium of parallel statement library |
| CN108829680A (en) * | 2018-06-22 | 2018-11-16 | 北京百悟科技有限公司 | A kind of violation publicity detection method and device, computer readable storage medium |
| CN109033390A (en) * | 2018-07-27 | 2018-12-18 | 深圳追科技有限公司 | The method and apparatus for automatically generating similar question sentence |
| CN109241251A (en) * | 2018-07-27 | 2019-01-18 | 众安信息技术服务有限公司 | A kind of session interaction method |
| CN109271492A (en) * | 2018-11-16 | 2019-01-25 | 广东小天才科技有限公司 | A kind of automatic generation method and system of corpus regular expression |
| CN109408821A (en) * | 2018-10-22 | 2019-03-01 | 腾讯科技（深圳）有限公司 | A kind of corpus generation method, calculates equipment and storage medium at device |
| CN109461039A (en) * | 2018-08-28 | 2019-03-12 | 厦门快商通信息技术有限公司 | A kind of text handling method and intelligent customer service method |
| CN109522534A (en) * | 2018-10-12 | 2019-03-26 | 北京来也网络科技有限公司 | Task creating method and device for corpus processing |
| CN109597873A (en) * | 2018-11-21 | 2019-04-09 | 腾讯科技（深圳）有限公司 | Processing method, device, computer-readable medium and the electronic equipment of corpus data |
| CN110196897A (en) * | 2019-05-23 | 2019-09-03 | 竹间智能科技（上海）有限公司 | A kind of case recognition methods based on question and answer template |
| CN110362803A (en) * | 2019-07-19 | 2019-10-22 | 北京邮电大学 | A kind of text template generation method based on the combination of domain features morphology |
| CN110727780A (en) * | 2019-10-17 | 2020-01-24 | 福建天晴数码有限公司 | System and method for automatically expanding acquaintance text |
| CN110738033A (en) * | 2018-07-03 | 2020-01-31 | 百度在线网络技术（北京）有限公司 | Report template generation method, device and storage medium |
| CN111274361A (en) * | 2020-01-21 | 2020-06-12 | 北京明略软件系统有限公司 | Industry new word discovery method and device, storage medium and electronic equipment |
| CN111309858A (en) * | 2020-01-20 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Information identification method, device, equipment and medium |
| CN111382256A (en) * | 2020-03-20 | 2020-07-07 | 北京百度网讯科技有限公司 | Information recommendation method and device |
| CN111552862A (en) * | 2019-12-28 | 2020-08-18 | 华南理工大学 | Automatic template mining system and method based on cross support degree evaluation |
| CN111597322A (en) * | 2019-12-28 | 2020-08-28 | 华南理工大学 | Automatic template mining system and method based on frequent item set |
| CN112948561A (en) * | 2021-03-29 | 2021-06-11 | 建信金融科技有限责任公司 | Method and device for automatically expanding question-answer knowledge base |
| CN113127610A (en) * | 2019-12-31 | 2021-07-16 | 北京猎户星空科技有限公司 | Data processing method, device, equipment and medium |
| CN113169931A (en) * | 2018-11-16 | 2021-07-23 | 利维帕尔森有限公司 | Script-based automated robot program creation |
| CN117130791A (en) * | 2023-10-26 | 2023-11-28 | 南通话时代信息科技有限公司 | Computing power resource allocation method and system of cloud customer service platform |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060004561A1 (en) * | 2004-06-30 | 2006-01-05 | Microsoft Corporation | Method and system for clustering using generalized sentence patterns |
| US20090106019A1 (en) * | 2004-08-31 | 2009-04-23 | Microsoft Corporation | Method and system for prioritizing communications based on sentence classifications |
| CN105868313A (en) * | 2016-03-25 | 2016-08-17 | 浙江大学 | Mapping knowledge domain questioning and answering system and method based on template matching technique |
| CN106649612A (en) * | 2016-11-29 | 2017-05-10 | 中国银联股份有限公司 | Method and device for matching automatic question and answer template |
-
2017
- 2017-12-26 CN CN201711436114.0A patent/CN108038234B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060004561A1 (en) * | 2004-06-30 | 2006-01-05 | Microsoft Corporation | Method and system for clustering using generalized sentence patterns |
| US20090106019A1 (en) * | 2004-08-31 | 2009-04-23 | Microsoft Corporation | Method and system for prioritizing communications based on sentence classifications |
| CN105868313A (en) * | 2016-03-25 | 2016-08-17 | 浙江大学 | Mapping knowledge domain questioning and answering system and method based on template matching technique |
| CN106649612A (en) * | 2016-11-29 | 2017-05-10 | 中国银联股份有限公司 | Method and device for matching automatic question and answer template |
Non-Patent Citations (1)
| Title |
|---|
| 向鑫: "基于语义逻辑推理的地理试题解答方法研究", 《中国优秀硕士学位论文全文数据库 哲学与人文科学辑》 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108776677A (en) * | 2018-05-28 | 2018-11-09 | 深圳前海微众银行股份有限公司 | Creation method, equipment and the computer readable storage medium of parallel statement library |
| CN108776677B (en) * | 2018-05-28 | 2021-11-12 | 深圳前海微众银行股份有限公司 | Parallel sentence library creating method and device and computer readable storage medium |
| CN108829680A (en) * | 2018-06-22 | 2018-11-16 | 北京百悟科技有限公司 | A kind of violation publicity detection method and device, computer readable storage medium |
| CN110738033A (en) * | 2018-07-03 | 2020-01-31 | 百度在线网络技术（北京）有限公司 | Report template generation method, device and storage medium |
| CN110738033B (en) * | 2018-07-03 | 2023-09-19 | 百度在线网络技术（北京）有限公司 | Report template generation method, device and storage medium |
| CN109241251A (en) * | 2018-07-27 | 2019-01-18 | 众安信息技术服务有限公司 | A kind of session interaction method |
| CN109241251B (en) * | 2018-07-27 | 2022-05-27 | 众安信息技术服务有限公司 | Conversation interaction method |
| CN109033390A (en) * | 2018-07-27 | 2018-12-18 | 深圳追科技有限公司 | The method and apparatus for automatically generating similar question sentence |
| CN109461039A (en) * | 2018-08-28 | 2019-03-12 | 厦门快商通信息技术有限公司 | A kind of text handling method and intelligent customer service method |
| CN109522534A (en) * | 2018-10-12 | 2019-03-26 | 北京来也网络科技有限公司 | Task creating method and device for corpus processing |
| CN109522534B (en) * | 2018-10-12 | 2022-12-13 | 北京来也网络科技有限公司 | Task generation method and device for corpus processing |
| CN109408821A (en) * | 2018-10-22 | 2019-03-01 | 腾讯科技（深圳）有限公司 | A kind of corpus generation method, calculates equipment and storage medium at device |
| CN109408821B (en) * | 2018-10-22 | 2020-09-04 | 腾讯科技（深圳）有限公司 | Corpus generation method and device, computing equipment and storage medium |
| CN109271492A (en) * | 2018-11-16 | 2019-01-25 | 广东小天才科技有限公司 | A kind of automatic generation method and system of corpus regular expression |
| CN113169931A (en) * | 2018-11-16 | 2021-07-23 | 利维帕尔森有限公司 | Script-based automated robot program creation |
| CN109597873A (en) * | 2018-11-21 | 2019-04-09 | 腾讯科技（深圳）有限公司 | Processing method, device, computer-readable medium and the electronic equipment of corpus data |
| CN109597873B (en) * | 2018-11-21 | 2022-02-08 | 腾讯科技（深圳）有限公司 | Corpus data processing method and device, computer readable medium and electronic equipment |
| CN110196897A (en) * | 2019-05-23 | 2019-09-03 | 竹间智能科技（上海）有限公司 | A kind of case recognition methods based on question and answer template |
| CN110362803A (en) * | 2019-07-19 | 2019-10-22 | 北京邮电大学 | A kind of text template generation method based on the combination of domain features morphology |
| CN110727780A (en) * | 2019-10-17 | 2020-01-24 | 福建天晴数码有限公司 | System and method for automatically expanding acquaintance text |
| CN111597322B (en) * | 2019-12-28 | 2023-04-21 | 华南理工大学 | Automatic template mining system and method based on frequent item sets |
| CN111552862B (en) * | 2019-12-28 | 2023-04-21 | 华南理工大学 | Automatic template mining system and method based on cross support evaluation |
| CN111552862A (en) * | 2019-12-28 | 2020-08-18 | 华南理工大学 | Automatic template mining system and method based on cross support degree evaluation |
| CN111597322A (en) * | 2019-12-28 | 2020-08-28 | 华南理工大学 | Automatic template mining system and method based on frequent item set |
| CN113127610A (en) * | 2019-12-31 | 2021-07-16 | 北京猎户星空科技有限公司 | Data processing method, device, equipment and medium |
| CN113127610B (en) * | 2019-12-31 | 2024-04-19 | 北京猎户星空科技有限公司 | Data processing method, device, equipment and medium |
| CN111309858A (en) * | 2020-01-20 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Information identification method, device, equipment and medium |
| CN111309858B (en) * | 2020-01-20 | 2023-03-07 | 腾讯科技（深圳）有限公司 | Information identification method, device, equipment and medium |
| CN111274361A (en) * | 2020-01-21 | 2020-06-12 | 北京明略软件系统有限公司 | Industry new word discovery method and device, storage medium and electronic equipment |
| CN111382256B (en) * | 2020-03-20 | 2024-04-09 | 北京百度网讯科技有限公司 | Information recommendation method and device |
| CN111382256A (en) * | 2020-03-20 | 2020-07-07 | 北京百度网讯科技有限公司 | Information recommendation method and device |
| CN112948561A (en) * | 2021-03-29 | 2021-06-11 | 建信金融科技有限责任公司 | Method and device for automatically expanding question-answer knowledge base |
| CN117130791B (en) * | 2023-10-26 | 2023-12-26 | 南通话时代信息科技有限公司 | Computing power resource allocation method and system of cloud customer service platform |
| CN117130791A (en) * | 2023-10-26 | 2023-11-28 | 南通话时代信息科技有限公司 | Computing power resource allocation method and system of cloud customer service platform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108038234B (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108038234A (en) | A kind of question sentence template automatic generation method and device | |
| CN106484664B (en) | Similarity calculating method between a kind of short text | |
| Bang et al. | Explaining a black-box by using a deep variational information bottleneck approach | |
| CN107368468A (en) | A kind of generation method and system of O＆M knowledge mapping | |
| CN108509519A (en) | World knowledge collection of illustrative plates enhancing question and answer interactive system based on deep learning and method | |
| CN105868184A (en) | Chinese name recognition method based on recurrent neural network | |
| CN106844346A (en) | Short text Semantic Similarity method of discrimination and system based on deep learning model Word2Vec | |
| CN112507136B (en) | Knowledge-driven business operation map construction method | |
| DE112013004082T5 (en) | Search system of the emotion entity for the microblog | |
| CN106503148B (en) | A kind of table entity link method based on multiple knowledge base | |
| CN107633005A (en) | A kind of knowledge mapping structure, comparison system and method based on class teaching content | |
| CN106570708A (en) | Management method and management system of intelligent customer service knowledge base | |
| CN108197294A (en) | A kind of text automatic generation method based on deep learning | |
| CN107301170A (en) | The method and apparatus of cutting sentence based on artificial intelligence | |
| CN112100397A (en) | Electric power plan knowledge graph construction method and system based on bidirectional gating circulation unit | |
| CN106886576A (en) | It is a kind of based on the short text keyword extracting method presorted and system | |
| CN106683667A (en) | Automatic rhythm extracting method, system and application thereof in natural language processing | |
| CN110162631A (en) | Chinese patent classification method, system and storage medium towards TRIZ inventive principle | |
| Khetarpal et al. | Spatial terms across languages support near-optimal communication: Evidence from Peruvian Amazonia, and computational analyses | |
| CN110188359A (en) | A kind of text entities abstracting method | |
| CN109670042A (en) | A kind of examination question classification and grade of difficulty method based on recurrent neural network | |
| CN111008215B (en) | Expert recommendation method combining label construction and community relation avoidance | |
| CN116010581A (en) | Knowledge graph question-answering method and system based on power grid hidden trouble shooting scene | |
| CN114265937A (en) | Intelligent classification analysis method and system of scientific and technological information, storage medium and server | |
| CN102063497A (en) | Open type knowledge sharing platform and entry processing method thereof |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240306 Address after: Room 1179, W Zone, 11th Floor, Building 1, No. 158 Shuanglian Road, Qingpu District, Shanghai, 201702 Patentee after: Shanghai Zhongan Information Technology Service Co.,Ltd. Country or region after: China Address before: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Patentee before: ZHONGAN INFORMATION TECHNOLOGY SERVICE Co.,Ltd. Country or region before: China |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240415 Address after: Room 1179, W Zone, 11th Floor, Building 1, No. 158 Shuanglian Road, Qingpu District, Shanghai, 201702 Patentee after: Shanghai Zhongan Information Technology Service Co.,Ltd. Country or region after: China Address before: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Patentee before: ZHONGAN INFORMATION TECHNOLOGY SERVICE Co.,Ltd. Country or region before: China |