CN108959531A - Information search method, device, equipment and storage medium - Google Patents
Information search method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN108959531A CN108959531A CN201810696192.2A CN201810696192A CN108959531A CN 108959531 A CN108959531 A CN 108959531A CN 201810696192 A CN201810696192 A CN 201810696192A CN 108959531 A CN108959531 A CN 108959531A
- Authority
- CN
- China
- Prior art keywords
- answer
- sentence
- question
- query statement
- class query
- 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
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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of information search method, device, equipment and storage mediums.This method comprises: determining that the answer type of question and answer class query statement of input is to be after non-class, the corresponding search result of acquisition question and answer class query statement;The sentence in described search result is extracted, determines the degree of correlation information of the sentence corresponding answer opinion classification and the sentence and the question and answer class query statement;For every kind of answer opinion classification, according to the degree of correlation information for the sentence and question and answer class query statement for belonging to current answer opinion classification, it is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as the corresponding viewpoint segment of current answer opinion classification, and determine the percent information of the corresponding search result of current answer opinion classification;Percent information and viewpoint segment are presented to user.Information search method provided in an embodiment of the present invention, answer type can be improved be non-class the corresponding search result of question and answer class query statement accuracy.
Description
Technical field
The present embodiments relate to technical field of information processing more particularly to a kind of information search method, device, equipment and
Storage medium.
Background technique
In internet when carrying out information search using search engine, inquiry (query) sentence of user's input can divide
For question and answer class query sentence and non-question and answer class query sentence.Wherein the answer type of question and answer class query sentence can be divided into reality
Body class describes class and is the classifications such as non-class.For example, its a kind of possible answer is the question and answer class query sentence for being non-class are as follows: pregnant
Woman can eat apple;Its answer are as follows: pregnant woman can eat apple.
Currently, being the question and answer class query sentence for being non-class for answer, question and answer are based on using the technology that viewpoint polymerize in advance
Problem viewpoint library is established to resource offline, is stored with viewpoint collection corresponding to a large amount of problem and problem in the problem viewpoint library
It closes, which includes the ratio of every kind of viewpoint, for example answer is the ratio of "Yes", the ratio that answer is "No", further includes
Support the web page contents of every kind of viewpoint.Receive on line user input be non-class question and answer class query sentence after, by looking into
Viewpoint library is inscribed in inquiry, can be obtained the ratio of the corresponding every kind of viewpoint of question and answer class query sentence, be supported the webpage of every kind of viewpoint
Content, and information will be obtained as search result and be presented to user.
The defect of above scheme is: on the one hand, the update in problem viewpoint library is slow, so that question and answer class query sentence pair
The accuracy for the search result answered is lower;On the other hand, the foundation in problem viewpoint library is based on question and answer to resource, so that question and answer class
The accuracy of the corresponding search result of query sentence also relies on question and answer to resource, the corresponding search knot of question and answer class query sentence
The accuracy of fruit is unable to get guarantee.
Summary of the invention
The embodiment of the present invention provides a kind of information search method, device, equipment and storage medium, and can be improved is that non-class is looked into
Ask the accuracy of the corresponding search result of sentence.
In a first aspect, the embodiment of the invention provides a kind of information search methods, this method comprises:
After determining that the answer type of question and answer class query statement of input is to be non-class, the question and answer class query statement is obtained
Corresponding search result;
The sentence in described search result is extracted, determines the corresponding answer opinion classification of the sentence and the sentence and institute
State the degree of correlation information of question and answer class query statement;
For every kind of answer opinion classification, language is inquired according to the sentence and the question and answer class that belong to current answer opinion classification
The degree of correlation information of sentence, is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as current answer viewpoint
Classify corresponding viewpoint segment, and determines the percent information of the corresponding search result of current answer opinion classification;
The percent information and the viewpoint segment are presented to user.
Second aspect, the embodiment of the invention also provides a kind of information search device, which includes:
Search result obtains module, for after determining that the answer type of question and answer class query statement of input is to be non-class,
Obtain the corresponding search result of the question and answer class query statement;
Degree of correlation information determination module determines that the sentence is corresponding and answers for extracting the sentence in described search result
The degree of correlation information of case opinion classification and the sentence and the question and answer class query statement;
Viewpoint segment determining module, for for every kind of answer opinion classification, according to belonging to current answer opinion classification
The degree of correlation information of sentence and the question and answer class query statement is subordinated in the sentence of current answer opinion classification and chooses at least one
Sentence determines the corresponding search result of current answer opinion classification as the corresponding viewpoint segment of current answer opinion classification
Percent information;
Viewpoint fragment display module, for the percent information and the viewpoint segment to be presented to user.
The third aspect the embodiment of the invention also provides a kind of computer equipment, including memory, processor and is stored in
On memory and the computer program that can run on a processor, the processor are realized when executing described program as the present invention is real
Apply method described in example.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, the program realize method as described in the embodiments of the present invention when being executed by processor.
The embodiment of the present invention obtains first after determining that the answer type of question and answer class query statement of input is to be non-class
Then the corresponding search result of question and answer class query statement extracts the sentence in search result, determines the corresponding answer viewpoint of sentence
The degree of correlation information of classification and sentence and question and answer class query statement is worked as subsequently for every kind of answer opinion classification according to belonging to
The sentence of preceding answer opinion classification and the degree of correlation information of question and answer class query statement, are subordinated to the sentence of current answer opinion classification
At least one sentence of middle selection determines current answer opinion classification pair as the corresponding viewpoint segment of current answer opinion classification
Percent information and viewpoint segment are finally presented to user by the percent information for the search result answered.It is provided in an embodiment of the present invention
Information search method determines that answer type is yes according to the degree of correlation information of the sentence of search result and question and answer class query statement
The corresponding viewpoint segment of each answer opinion classification of the question and answer class query statement of non-class, and determine every kind of answer opinion classification pair
The ratio for the search result answered, answer type can be improved be non-class the corresponding search result of question and answer class query statement standard
True property.
Detailed description of the invention
Fig. 1 is the flow chart of one of the embodiment of the present invention one information search method;
Fig. 2 is the flow chart of one of the embodiment of the present invention two information search method;
Fig. 3 is the flow chart of one of the embodiment of the present invention three information search method;
Fig. 4 is the flow chart of another information search method in the embodiment of the present invention three;
Fig. 5 is the structural schematic diagram of one of the embodiment of the present invention four information search device.
Fig. 6 is the structural schematic diagram of one of the embodiment of the present invention five computer equipment.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is the flow chart of a kind of information search method that the embodiment of the present invention one provides, and the present embodiment is applicable to pair
Answer type be non-class question and answer class query statement carry out information search the case where, this method can by information search device Lai
It executes, which can be made of hardware and/or software, and can generally be integrated in computer, server and all search comprising information
In the terminal of Suo Gongneng.As shown in Figure 1, this method specifically comprises the following steps.
Step 110, determining that the answer type of question and answer class query statement of input is to be after non-class, acquisition question and answer class inquiry
The corresponding search result of sentence.
The answer type of question and answer class query statement includes entity class, describes class and is non-class etc..It wherein, is non-class inquiry
Sentence can be the query statement that answer includes different viewpoints classification, for example, it is that " tomato is nice that one, which is non-class query statement,
? ", answer may include " nice ", " not being very good eating " and " neutrality " these three viewpoints.Search result can be search engine
For input question and answer class query statement feedback as a result, search result can be the form of webpage, each search result can be with
The link etc. of title and search result web page comprising search result web page.Search engine can be Baidu etc..
Specifically, first determining whether that question and answer class is looked into after user inputs question and answer class query statement to the search box of search engine
Ask sentence answer type whether be non-class, if being non-class, then continue to obtain search engine for question and answer class query statement
The search result of feedback does not continue any operation then if not being non-class.
Step 120, the sentence in search result is extracted, determines the corresponding answer opinion classification of sentence and sentence and question and answer class
The degree of correlation information of query statement.
Wherein, the sentence in search result can be sentence or search result web page in the title of search result web page
Document in sentence.Answer opinion classification may include positive viewpoint, neutral viewpoint and negative sense viewpoint.Sentence and question and answer class
The degree of correlation information of query statement may include that sentence is aligned search where probability value, sentence with the viewpoint of question and answer class query statement
As a result the semantic similarity for matching angle value and sentence and question and answer class query statement of web page title and question and answer class query statement
At least one of value.
Optionally, it determines that the mode of the corresponding answer opinion classification of sentence can be, text emotion is passed through to each sentence
The corresponding answer opinion classification of tendency extracting tool extraction (including positive viewpoint, negative sense viewpoint, neutral viewpoint are answered;Alternatively, to mentioning
Sentence in the search result got carries out semantic analysis, obtains the corresponding answer opinion classification of sentence.
Optionally, determine that sentence is aligned probability value with the viewpoint of question and answer class query statement and can implement by following manner: root
According to the alignment probability matrix pre-established, being aligned generally for the focus word and each word in sentence in question and answer class query statement is determined
Rate value;Probability value is aligned using maximum alignment probability value as sentence with the viewpoint of question and answer class query statement.
Wherein, alignment probability matrix can be by the focus word and search result in question and answer class query statement in sentence
The matrix of the alignment probability value composition of each word.It is exemplary comprising descriptor, interrogative and focus word in question and answer class query statement
, for question and answer class query statement " tomato is nice ", " tomato " is descriptor, and " " is interrogative, and " nice " is burnt
Point is eaten.
Specifically, the word for including in sentence is separated and is obtained first after obtaining the sentence in search result
Then focus word in question and answer class query statement is searched each in focus word and sentence in the alignment probability matrix pre-established
The alignment probability value of word is finally aligned probability using maximum alignment probability value as sentence with the viewpoint of question and answer class query statement
Value.Illustratively, for " tomato is nice ", the sentence of one of search result is " too sour, not to be very good eating ", sentence packet
The word contained has " too sour " and " not being very good eating ", wherein the probability value that is aligned of " too sour " and focus word is 8%, " not being very good eating " with
The alignment probability value of focus word is 10%, then " too sour, not to be very good eating " is aligned probability value with the viewpoint of " tomato be fond of eating " and is
10%.
Optionally, according to the alignment probability matrix pre-established, the focus word and language in question and answer class query statement are determined
Further include following steps before the alignment probability value of each word in sentence: obtaining the identical multiple random question and answer classes of focus word and look into
Ask the corresponding search result of sentence;Count the frequency values that each word occurs in search result, the frequency values that statistics is obtained as
The focus word of random question and answer class query statement is aligned probability value with equivalent, is stored in alignment probability matrix.
Specifically, the identical multiple random question and answer class query statements of focus word are inputted search engine respectively, obtain each
The corresponding search result of question and answer class query statement at random, extracts the word in the sentence of each search result, and count each word
The frequency values occurred in all search results, the frequency values for finally obtaining statistics are as the coke of random question and answer class query statement
The probability value that is aligned of point word and equivalent is stored in alignment probability matrix.Illustratively, the focus word of " tomato is nice "
It is " nice ", answers the corresponding search result of class query statement at random comprising the multiple of " nice " then obtaining, and counts search knot
The frequency values that each word occurs in fruit sentence are aligned probability value using frequency values that statistics obtains as " nice " and equivalent,
It is stored in alignment probability matrix.For example, in search result sentence some word occur frequency values i.e. the word occur number with
The ratio of the total number of the included word of search result is 80%, then " nice " is 80% with the probability value that is aligned of the word.
Optionally, the matching angle value of the web page title of search result and question and answer class query statement, can lead to where determining sentence
Crossing following manner is to implement: the web page title input content matching degree of search result where question and answer class query statement and sentence is true
Cover half type obtains content matching degree and determines the sentence of model output and the matching angle value of question and answer class query statement.
Wherein, matching degree can use the click volume of the webpage of sentence place search result to characterize.Content matching degree is true
Cover half type can be based on the first training data source, using deep neural network (Deep Neural Network, DNN) algorithm
It is trained the model of acquisition.Wherein, the first training data source may include: a plurality of random question and answer class query statement and every with
Machine question and answer class query statement it is corresponding by user click the search result from question and answer station Web page subject and by user's point
The click volume of the search result from question and answer station hit.In the present embodiment, search where question and answer class query statement and sentence is tied
After the web page title input content matching degree of fruit determines model, content matching degree model is to where question and answer class query statement and sentence
The web page title of search result is analyzed, and the matching angle value of sentence and question and answer class query statement is obtained.
Optionally, determine that sentence and the mode of the semantic similarity value of question and answer class query statement can be, by sentence with ask
Class query statement input semantic similarity model is answered, the semantic similarity value of semantic similarity model output is obtained.Wherein, semantic
Similarity model can be based on training data source, match (Deep Structured Semantic using deep semantic
Models, DSSM) algorithm is trained the model of acquisition.Wherein, training data source may include multi-group data pair, every group of data
To including random question and answer class query statement, the sentence in search result corresponding with the random question and answer class query statement, Yi Jili
With the semantic similarity value of random question and answer quasi-sentence and the sentence in search result that cosine similarity algorithm calculates.
Step 130, it for every kind of answer opinion classification, is looked into according to the sentence for belonging to current answer opinion classification with question and answer class
The degree of correlation information for asking sentence, is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as current answer
The corresponding viewpoint segment of opinion classification, and determine the percent information of the corresponding search result of current answer opinion classification.
Optionally, believed according to the degree of correlation of the sentence and the question and answer class query statement that belong to current answer opinion classification
Breath, is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as the corresponding sight of current answer opinion classification
Point segment, can be implemented: the sentence for belonging to current answer opinion classification is related to question and answer class query statement by following manner
Information is spent, the disaggregated model pre-established is inputted;Obtain the probability value that the correspondence sentence of disaggregated model output is selected;By probability
It is worth highest sentence as the corresponding viewpoint segment of current answer opinion classification.
Wherein, disaggregated model can be based on the second training data source, using support vector machines (Support Vector
Machine, SVM) algorithm is trained the model of acquisition.Second training data source may include multi-group data pair, every group of data
To including random question and answer class query statement, the sentence in search result corresponding with the random question and answer class query statement, the sentence
The markup information of viewpoint segment whether is selected as with the degree of correlation information of the random question and answer class query statement and the sentence.This
In embodiment, after obtaining the sentence of current answer opinion classification and the degree of correlation information of question and answer class query statement, the degree of correlation is believed
In breath input disaggregated model, disaggregated model analyzes degree of correlation information, obtains the probability value that corresponding sentence is selected, finally
By the corresponding viewpoint segment of the maximum sentence of probability value answer opinion classification the most current.Illustratively, for " tomato is nice
", the sentence of the search result of positive viewpoint answer have 3 " good, vitamin is very high ", " acid acid, mouthfeel is fine " and
This 3 sentences and the degree of correlation information of " tomato is nice " are inputted disaggregated model respectively, obtained by " nice, nutritive value is high "
" good, vitamin is very high " probability value for being selected be 25%, the probability value that " acid sour, mouthfeel is fine " is selected for
40%, the probability that " nice, nutritive value is high " is selected is 35%, then answers " acid acid, mouthfeel is fine " as positive viewpoint
The corresponding viewpoint segment of case.
Optionally, determine that the mode of the percent information of the corresponding search result of current answer opinion classification can be, according to
The quantity of the sum search result corresponding with current answer opinion classification of search result calculates percent information.Such as: certain question and answer
The search result of class query statement has 100, wherein positive viewpoint answer has 60, negative sense viewpoint answer has 30, and neutrality is seen
Point answer has 10, then the percent information of the corresponding search result of positive viewpoint answer is 60%, and negative sense viewpoint answer is corresponding
The percent information of search result is 30%, and the percent information of the corresponding search result of neutral viewpoint answer is 10%.
Step 140, percent information and viewpoint segment are presented to user.
Specifically, will compare after obtaining the percent information of viewpoint segment and the corresponding search result of answer opinion classification
Example information and viewpoint segment are presented to user.The mode showed can be to be shown on the top of result of page searching.
The technical solution of the present embodiment is determining that the answer type of question and answer class query statement of input is to be after non-class, head
The corresponding search result of question and answer class query statement is first obtained, the sentence in search result is then extracted, determines that sentence is corresponding and answer
The degree of correlation information of case opinion classification and sentence and question and answer class query statement, subsequently for every kind of answer opinion classification, according to
Belong to the sentence of current answer opinion classification and the degree of correlation information of question and answer class query statement, is subordinated to current answer opinion classification
Sentence in choose at least one sentence as the corresponding viewpoint segment of current answer opinion classification, and determine current answer viewpoint
Classify the percent information of corresponding search result, percent information and viewpoint segment are finally presented to user.The embodiment of the present invention
The information search method of offer determines answer class according to the degree of correlation information of the sentence of search result and question and answer class query statement
Type be non-class question and answer class query statement the corresponding viewpoint segment of each answer opinion classification, and determine every kind of answer viewpoint
Classify the ratio of corresponding search result, it is the corresponding search knot of question and answer class query statement for being non-class that answer type, which can be improved,
The accuracy of fruit.
Embodiment two
Fig. 2 is a kind of flow chart of information search method provided by Embodiment 2 of the present invention, based on above-described embodiment,
As shown in Fig. 2, this method comprises the following steps.
Step 210, the first training data source is obtained.
Wherein, the first training data source includes: a plurality of random question and answer class query statement and every random question and answer class inquiry language
Sentence it is corresponding by user click the search result from question and answer station Web page subject and by user click come from question and answer station
Search result click volume.
The mode for obtaining the first training data source is to can be, and it is non-class that a plurality of answer type is obtained in question and answer station to be
The webpage master of random question and answer class query statement and the corresponding search result clicked by user of every random question and answer class query statement
The click volume of topic and the search result from question and answer station clicked by user, and the amount of will click on is labeled as search result
The matching angle value of sentence and question and answer class query statement.
Step 220, it is based on the first training data source, model training is carried out using DNN algorithm, content matching degree is obtained and determines
Model.
Specifically, model training is constantly carried out using DNN algorithm, in training process behind the first training data source of acquisition
In, the parameter in DNN algorithm is constantly adjusted, until model has the ability of accurate output matching angle value, to obtain content
Model is determined with spending.
Step 230, the web page title input content matching degree of search result where question and answer class query statement and sentence is true
Cover half type obtains content matching degree and determines the sentence of model output and the matching angle value of question and answer class query statement.
The technical solution of the present embodiment is based on the first training data source, carries out model training using DNN algorithm, obtains interior
Hold matching degree and determines model.Search result where content matching degree model determines question and answer class query statement and sentence can be improved
The accuracy of web page title matching angle value.
Embodiment three
Fig. 3 is a kind of flow chart for information search method that the embodiment of the present invention three provides, based on above-described embodiment,
As shown in figure 3, this method comprises the following steps
Step 310, the second training data source is obtained.
Wherein, the second training data source includes: multi-group data pair, every group of data to include random question and answer class query statement,
The phase of sentence, the sentence and the random question and answer class query statement in search result corresponding with the random question and answer class query statement
Whether pass degree information and the sentence are selected as the markup information of viewpoint segment.
Specifically, being labeled as not being selected as viewpoint segment to sentence if sentence is not selected as viewpoint segment, if sentence quilt
It is elected to be viewpoint segment, then sentence is labeled as being selected as viewpoint segment.In the present embodiment, the mode in the second training data source is obtained
It can be, obtaining a plurality of answer type in the question and answer station of search engine is the random question and answer class query statement for being non-class, every
Sentence in the corresponding search result of question and answer class query statement at random, and obtain the sentence in search result and looked into random question and answer class
Whether the degree of correlation information and sentence of asking sentence are selected as the markup information of viewpoint segment.
Step 320, it is based on the second training data source, model training is carried out using SVM algorithm, obtains disaggregated model.
Specifically, model training is constantly carried out using SVM algorithm, in training process behind the second training data source of acquisition
In, the parameter in SVM algorithm is constantly adjusted, until model has the ability of accurate output probability value, to obtain disaggregated model.
Step 330, the degree of correlation information of the sentence and question and answer class query statement of current answer opinion classification will be belonged to, inputted
The disaggregated model pre-established obtains the probability value that the correspondence sentence of disaggregated model output is selected.
Step 340, using the highest sentence of probability value as the corresponding viewpoint segment of current answer opinion classification.
The technical solution of the present embodiment is based on the second training data source, carries out model training using SVM algorithm, is divided
Class model.The accuracy for the probability value that disaggregated model determines that sentence is selected can be improved.
Fig. 4 is the flow chart for another information search method that the embodiment of the present invention three provides, as to above-described embodiment
Be explained further, as shown in figure 4, this method comprises the following steps.
Step 410, determining that the answer type of question and answer class query statement of input is to be after non-class, acquisition question and answer class inquiry
The corresponding search result of sentence.
Step 420, the sentence in search result is extracted, determines the corresponding answer opinion classification of sentence.
Step 430, determine that sentence and the viewpoint of question and answer class query statement are aligned probability value.
Step 440, the matching angle value of the web page title of search result and question and answer class query statement where determining sentence.
Step 450, the semantic similarity value of sentence Yu question and answer class query statement is determined.
Step 460, it for every kind of answer opinion classification, is looked into according to the sentence for belonging to current answer opinion classification with question and answer class
The degree of correlation information for asking sentence, is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as current answer
The corresponding viewpoint segment of opinion classification.
Step 470, the percent information of the corresponding search result of answer opinion classification is determined.
Step 480, percent information and viewpoint segment are presented to user.
Example IV
Fig. 5 is a kind of structural schematic diagram for information search device that the embodiment of the present invention four provides.As shown in figure 5, the dress
Set includes: that search result obtains module 510, degree of correlation information determination module 520, viewpoint segment determining module 530 and viewpoint piece
Section display module 540.
Search result obtains module 510, is non-class in the answer type for determining the question and answer class query statement of input for being
Afterwards, the corresponding search result of question and answer class query statement is obtained;
Degree of correlation information determination module 520 determines the corresponding answer viewpoint of sentence for extracting the sentence in search result
The degree of correlation information of classification and sentence and question and answer class query statement;
Viewpoint segment determining module 530, for for every kind of answer opinion classification, according to belonging to current answer opinion classification
Sentence and question and answer class query statement degree of correlation information, be subordinated in the sentence of current answer opinion classification and choose at least one
Sentence determines the corresponding search result of current answer opinion classification as the corresponding viewpoint segment of current answer opinion classification
Percent information;
Viewpoint fragment display module 540, for percent information and viewpoint segment to be presented to user.
Optionally, the degree of correlation information of sentence and question and answer class query statement includes:
The web page title and question and answer of search result where sentence is aligned probability value, sentence with the viewpoint of question and answer class query statement
The matching angle value and sentence of class query statement and at least one of the semantic similarity value of question and answer class query statement.
Optionally, degree of correlation information determination module 520, is also used to:
According to the alignment probability matrix pre-established, each of focus word in question and answer class query statement and sentence are determined
The alignment probability value of word;
Probability value is aligned using maximum alignment probability value as sentence with the viewpoint of question and answer class query statement.
Optionally, further includes:
Search result acquisition submodule, for obtaining, the identical multiple random question and answer class query statements of focus word are corresponding to be searched
Hitch fruit;
It is aligned probability matrix and obtains module, for counting the frequency values that each word occurs in search result, statistics is obtained
Frequency values be aligned probability value as focus word and the equivalent of random question and answer class query statement, be stored in alignment probability matrix
In.
Optionally, degree of correlation information determination module 520, is also used to:
The web page title input content matching degree of search result where question and answer class query statement and sentence is determined into model, is obtained
The sentence of model output and the matching angle value of question and answer class query statement are determined to content matching degree.
Optionally, further includes:
First training data source obtains module, and for obtaining the first training data source, the first training data source includes: a plurality of
Random question and answer class query statement and the corresponding search from question and answer station clicked by user of every random question and answer class query statement
As a result the click volume of Web page subject and the search result from question and answer station clicked by user;
Content matching degree determines that model obtains module, for being based on the first training data source, using deep neural network DNN
Algorithm carries out model training, obtains content matching degree and determines model.
Viewpoint segment determining module 530, is also used to:
The degree of correlation information of the sentence and question and answer class query statement of current answer opinion classification will be belonged to, input pre-establishes
Disaggregated model;Obtain the probability value that the correspondence sentence of disaggregated model output is selected;
Using the highest sentence of probability value as the corresponding viewpoint segment of current answer opinion classification.
Optionally, further includes:
Second training data source obtains module, and for obtaining the second training data source, the second training data source includes: multiple groups
Data pair, every group of data are to including random question and answer class query statement, search result corresponding with the random question and answer class query statement
In sentence, the sentence and the random question and answer class query statement degree of correlation information and the sentence whether be selected as viewpoint piece
The markup information of section;
Disaggregated model obtains module, for being based on the second training data source, carries out model using support vector machines algorithm
Training, obtains disaggregated model.
Method provided by the executable aforementioned all embodiments of the present invention of above-mentioned apparatus, it is corresponding to have the execution above method
Functional module and beneficial effect.The not technical detail of detailed description in the present embodiment, reference can be made to the aforementioned all implementations of the present invention
Method provided by example.
Embodiment five
Fig. 6 is a kind of structural schematic diagram for computer equipment that the embodiment of the present invention five provides, as shown in fig. 6, this implementation
A kind of computer equipment that example provides, comprising: processor 61 and memory 62.Processor in the computer equipment can be one
A or multiple, in Fig. 6 by taking a processor 61 as an example, processor 61 and memory 62 in the computer equipment can pass through
Bus or other modes connect, in Fig. 6 for being connected by bus.
Information search device provided by the above embodiment is integrated in the processor 61 of computer equipment in the present embodiment.This
Outside, the memory 62 in the computer equipment is used as a kind of computer readable storage medium, can be used for storing one or more journeys
Sequence, described program can be software program, computer executable program and module, such as information search side in the embodiment of the present invention
Corresponding program instruction/the module of method.Software program, instruction and the mould that processor 61 is stored in memory 62 by operation
Block, thereby executing the various function application and data processing of equipment, i.e. information search method in realization above method embodiment.
Memory 62 may include storing program area and storage data area, wherein storing program area can storage program area, extremely
Application program needed for a few function;Storage data area, which can be stored, uses created data etc. according to equipment.In addition, depositing
Reservoir 62 may include high-speed random access memory, can also include nonvolatile memory, and a for example, at least disk is deposited
Memory device, flush memory device or other non-volatile solid state memory parts.In some instances, memory 62 can further comprise
The memory remotely located relative to processor 61, these remote memories can pass through network connection to equipment.Above-mentioned network
Example include but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
The program that processor 61 is stored in memory 62 by operation, at various function application and data
Reason realizes example information search method provided in an embodiment of the present invention.
Embodiment six
The embodiment of the present invention six additionally provides a kind of computer readable storage medium, is stored thereon with computer program, should
The information search method as provided by the embodiment of the present invention is realized when program is executed by processor.
Certainly, a kind of computer readable storage medium provided by the embodiment of the present invention, the computer program stored thereon
The method operation being not limited to the described above, can also be performed the phase in information search method provided by any embodiment of the invention
Close operation.
The computer storage medium of the embodiment of the present invention, can be using any of one or more computer-readable media
Combination.Computer-readable medium can be computer-readable signal media or computer readable storage medium.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example (non exhaustive list) of computer readable storage medium includes: tool
There are electrical connection, the portable computer diskette, hard disk, random access memory (RAM), read-only memory of one or more conducting wires
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable storage
Medium can be any tangible medium for including or store program, which can be commanded execution system, device or device
Using or it is in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
It further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on a remote computer or server completely on the remote computer on the user computer.?
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as mentioned using Internet service
It is connected for quotient by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (11)
1. a kind of information search method characterized by comprising
After determining that the answer type of question and answer class query statement of input is to be non-class, it is corresponding to obtain the question and answer class query statement
Search result;
The sentence in described search result is extracted, determines that the corresponding answer opinion classification of the sentence and the sentence are asked with described
Answer the degree of correlation information of class query statement;
For every kind of answer opinion classification, according to the sentence and the question and answer class query statement for belonging to current answer opinion classification
Degree of correlation information is subordinated in the sentence of current answer opinion classification and chooses at least one sentence as current answer opinion classification
Corresponding viewpoint segment, and determine the percent information of the corresponding search result of current answer opinion classification;
The percent information and the viewpoint segment are presented to user.
2. the method according to claim 1, wherein the degree of correlation of the sentence and the question and answer class query statement
Information includes:
The webpage mark of search result where the sentence is aligned probability value, the sentence with the viewpoint of the question and answer class query statement
It inscribes and the semantic similarity for matching angle value and the sentence and the question and answer class query statement of the question and answer class query statement
At least one of value.
3. according to the method described in claim 2, it is characterized in that, determining the sight of the sentence and the question and answer class query statement
Point alignment probability value, comprising:
According to the alignment probability matrix pre-established, determine in the focus word and the sentence in the question and answer class query statement
The alignment probability value of each word;
Probability value is aligned using maximum alignment probability value as the sentence with the viewpoint of the question and answer class query statement.
4. according to the method described in claim 3, it is characterized in that, determining institute according to the alignment probability matrix pre-established
State being aligned before probability value for focus word in question and answer class query statement and each word in the sentence, further includes:
Obtain the corresponding search result of the identical multiple random question and answer class query statements of focus word;
The frequency values that each word occurs in described search result are counted, the frequency values that statistics is obtained are as the random question and answer class
The focus word of query statement is aligned probability value with equivalent, is stored in alignment probability matrix.
5. according to the method described in claim 2, it is characterized in that, the web page title of search result where determining the sentence with
The matching angle value of the question and answer class query statement, comprising:
The web page title input content matching degree of search result where the question and answer class query statement and the sentence is determined into mould
Type obtains content matching degree and determines the sentence of model output and the matching angle value of the question and answer class query statement.
6. according to the method described in claim 5, it is characterized in that, where by the question and answer class query statement and the sentence
The web page title input content matching degree of search result determines before model, further includes:
Obtain the first training data source, first training data source include: a plurality of random question and answer class query statement and every with
Machine question and answer class query statement it is corresponding by user click the search result from question and answer station Web page subject and it is described by with
The click volume for the search result from question and answer station that family is clicked;
Based on first training data source, model training is carried out using deep neural network DNN algorithm, obtains the content
Model is determined with spending.
7. the method according to claim 1, wherein according to belong to the sentence of current answer opinion classification with it is described
The degree of correlation information of question and answer class query statement is subordinated in the sentence of current answer opinion classification and chooses at least one sentence conduct
The corresponding viewpoint segment of currently answer opinion classification, comprising:
The degree of correlation information of the sentence and the question and answer class query statement of current answer opinion classification will be belonged to, input pre-establishes
Disaggregated model;Obtain the probability value that the correspondence sentence of the disaggregated model output is selected;
Using the highest sentence of probability value as the corresponding viewpoint segment of current answer opinion classification.
8. the method according to the description of claim 7 is characterized in that the sentence that will belong to current answer opinion classification with it is described
The degree of correlation information of question and answer class query statement, inputs before the disaggregated model pre-established, further includes:
The second training data source is obtained, second training data source includes: multi-group data pair, and every group of data to including asking at random
Answer class query statement, sentence, the sentence and the random question and answer in search result corresponding with the random question and answer class query statement
Whether the degree of correlation information of class query statement and the sentence are selected as the markup information of viewpoint segment;
Based on second training data source, model training is carried out using support vector machines algorithm, obtains the classification mould
Type.
9. a kind of information search device characterized by comprising
Search result obtains module, for determining that the answer type of question and answer class query statement of input is to be after non-class, acquisition
The corresponding search result of the question and answer class query statement;
Degree of correlation information determination module determines that the corresponding answer of the sentence is seen for extracting the sentence in described search result
The degree of correlation information of point classification and the sentence and the question and answer class query statement;
Viewpoint segment determining module is used for for every kind of answer opinion classification, according to the sentence for belonging to current answer opinion classification
With the degree of correlation information of the question and answer class query statement, it is subordinated in the sentence of current answer opinion classification and chooses at least one language
Sentence determines the ratio of the corresponding search result of current answer opinion classification as the corresponding viewpoint segment of current answer opinion classification
Example information;
Viewpoint fragment display module, for the percent information and the viewpoint segment to be presented to user.
10. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, which is characterized in that the processor realizes such as side described in any one of claims 1-8 when executing described program
Method.
11. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Such as method described in any one of claims 1-8 is realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810696192.2A CN108959531B (en) | 2018-06-29 | 2018-06-29 | Information searching method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810696192.2A CN108959531B (en) | 2018-06-29 | 2018-06-29 | Information searching method, device, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108959531A true CN108959531A (en) | 2018-12-07 |
CN108959531B CN108959531B (en) | 2020-12-22 |
Family
ID=64488084
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810696192.2A Active CN108959531B (en) | 2018-06-29 | 2018-06-29 | Information searching method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108959531B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635197A (en) * | 2018-12-17 | 2019-04-16 | 北京百度网讯科技有限公司 | Searching method, device, electronic equipment and storage medium |
CN110059190A (en) * | 2019-04-18 | 2019-07-26 | 东南大学 | A kind of user's real-time point of view detection method based on social media content and structure |
CN110209786A (en) * | 2019-05-28 | 2019-09-06 | 北京百度网讯科技有限公司 | It is display methods, device, computer equipment and the storage medium of non-class answer |
CN110297954A (en) * | 2019-06-14 | 2019-10-01 | 北京百度网讯科技有限公司 | Answer verifying method, device, equipment and storage medium |
CN111125334A (en) * | 2019-12-20 | 2020-05-08 | 神思电子技术股份有限公司 | Search question-answering system based on pre-training |
CN111858880A (en) * | 2020-06-18 | 2020-10-30 | 北京百度网讯科技有限公司 | Method and device for obtaining query result, electronic equipment and readable storage medium |
CN112100345A (en) * | 2020-08-25 | 2020-12-18 | 百度在线网络技术(北京)有限公司 | Training method and device for non-question-answer-like model, electronic equipment and storage medium |
CN112395314A (en) * | 2019-08-16 | 2021-02-23 | 北京大数医达科技有限公司 | Method, electronic device and computer readable medium for searching information |
CN112860865A (en) * | 2021-02-10 | 2021-05-28 | 达而观信息科技(上海)有限公司 | Method, device, equipment and storage medium for realizing intelligent question answering |
CN113392308A (en) * | 2021-06-22 | 2021-09-14 | 北京字节跳动网络技术有限公司 | Content search method, device, equipment and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927381A (en) * | 2014-04-29 | 2014-07-16 | 北京百度网讯科技有限公司 | Right-and-wrong problem processing method and device |
CN105718543A (en) * | 2016-01-18 | 2016-06-29 | 腾讯科技(深圳)有限公司 | Sentence display method and device |
CN106682194A (en) * | 2016-12-29 | 2017-05-17 | 北京百度网讯科技有限公司 | Answer positioning method and device based on deep questions and answers |
US20170293680A1 (en) * | 2016-04-06 | 2017-10-12 | International Business Machines Corporation | Natural language processing based on textual polarity |
-
2018
- 2018-06-29 CN CN201810696192.2A patent/CN108959531B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927381A (en) * | 2014-04-29 | 2014-07-16 | 北京百度网讯科技有限公司 | Right-and-wrong problem processing method and device |
CN105718543A (en) * | 2016-01-18 | 2016-06-29 | 腾讯科技(深圳)有限公司 | Sentence display method and device |
US20170293680A1 (en) * | 2016-04-06 | 2017-10-12 | International Business Machines Corporation | Natural language processing based on textual polarity |
CN106682194A (en) * | 2016-12-29 | 2017-05-17 | 北京百度网讯科技有限公司 | Answer positioning method and device based on deep questions and answers |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635197B (en) * | 2018-12-17 | 2021-08-24 | 北京百度网讯科技有限公司 | Searching method, searching device, electronic equipment and storage medium |
CN109635197A (en) * | 2018-12-17 | 2019-04-16 | 北京百度网讯科技有限公司 | Searching method, device, electronic equipment and storage medium |
US11709893B2 (en) | 2018-12-17 | 2023-07-25 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Search method, electronic device and storage medium |
CN110059190A (en) * | 2019-04-18 | 2019-07-26 | 东南大学 | A kind of user's real-time point of view detection method based on social media content and structure |
CN110209786A (en) * | 2019-05-28 | 2019-09-06 | 北京百度网讯科技有限公司 | It is display methods, device, computer equipment and the storage medium of non-class answer |
CN110297954A (en) * | 2019-06-14 | 2019-10-01 | 北京百度网讯科技有限公司 | Answer verifying method, device, equipment and storage medium |
CN110297954B (en) * | 2019-06-14 | 2022-08-30 | 北京百度网讯科技有限公司 | Answer verification processing method, device, equipment and storage medium |
CN112395314A (en) * | 2019-08-16 | 2021-02-23 | 北京大数医达科技有限公司 | Method, electronic device and computer readable medium for searching information |
CN111125334A (en) * | 2019-12-20 | 2020-05-08 | 神思电子技术股份有限公司 | Search question-answering system based on pre-training |
CN111125334B (en) * | 2019-12-20 | 2023-09-12 | 神思电子技术股份有限公司 | Search question-answering system based on pre-training |
CN111858880A (en) * | 2020-06-18 | 2020-10-30 | 北京百度网讯科技有限公司 | Method and device for obtaining query result, electronic equipment and readable storage medium |
CN111858880B (en) * | 2020-06-18 | 2024-01-26 | 北京百度网讯科技有限公司 | Method, device, electronic equipment and readable storage medium for obtaining query result |
CN112100345A (en) * | 2020-08-25 | 2020-12-18 | 百度在线网络技术(北京)有限公司 | Training method and device for non-question-answer-like model, electronic equipment and storage medium |
CN112860865A (en) * | 2021-02-10 | 2021-05-28 | 达而观信息科技(上海)有限公司 | Method, device, equipment and storage medium for realizing intelligent question answering |
CN113392308A (en) * | 2021-06-22 | 2021-09-14 | 北京字节跳动网络技术有限公司 | Content search method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN108959531B (en) | 2020-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108959531A (en) | Information search method, device, equipment and storage medium | |
US20200301954A1 (en) | Reply information obtaining method and apparatus | |
CN110209897B (en) | Intelligent dialogue method, device, storage medium and equipment | |
KR102288249B1 (en) | Information processing method, terminal, and computer storage medium | |
CN109325148A (en) | The method and apparatus for generating information | |
CN106407178A (en) | Session abstract generation method and device | |
CN107491547A (en) | Searching method and device based on artificial intelligence | |
CN107832432A (en) | A kind of search result ordering method, device, server and storage medium | |
CN110491383A (en) | A kind of voice interactive method, device, system, storage medium and processor | |
CN108228576B (en) | Text translation method and device | |
CN112799747A (en) | Intelligent assistant evaluation and recommendation method, system, terminal and readable storage medium | |
CN107679082A (en) | Question and answer searching method, device and electronic equipment | |
CN109918627B (en) | Text generation method, device, electronic equipment and storage medium | |
CN109800307A (en) | Analysis method, device, computer equipment and the storage medium of product evaluation | |
CN110597992A (en) | Semantic reasoning method and device based on knowledge graph and electronic equipment | |
CN109582954A (en) | Method and apparatus for output information | |
CN108960574A (en) | Quality determination method, device, server and the storage medium of question and answer | |
CN109992781B (en) | Text feature processing method and device and storage medium | |
CN113254711B (en) | Interactive image display method and device, computer equipment and storage medium | |
CN110019729B (en) | Intelligent question-answering method, storage medium and terminal | |
CN108959529A (en) | Determination method, apparatus, equipment and the storage medium of problem answers type | |
CN114090755A (en) | Reply sentence determination method and device based on knowledge graph and electronic equipment | |
CN107967304A (en) | Session interaction processing method, device and electronic equipment | |
CN110309293A (en) | Text recommended method and device | |
CN112667792B (en) | Man-machine dialogue data processing method and device, computer equipment and storage medium |
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 |