CN106776782A - Semantic similarity acquisition methods and device based on artificial intelligence - Google Patents
Semantic similarity acquisition methods and device based on artificial intelligence Download PDFInfo
- Publication number
- CN106776782A CN106776782A CN201611042515.3A CN201611042515A CN106776782A CN 106776782 A CN106776782 A CN 106776782A CN 201611042515 A CN201611042515 A CN 201611042515A CN 106776782 A CN106776782 A CN 106776782A
- Authority
- CN
- China
- Prior art keywords
- grain size
- search
- size characteristic
- weight
- entry
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention proposes a kind of semantic similarity acquisition methods and device based on artificial intelligence, wherein, method includes：By after many grain size characteristics for obtaining query and title, obtain the weight of each grain size characteristic, the significance level that varigrained feature has can be embodied by the weight, then when representing multiple graininess is carried out to query and title, add the weight of each grain size characteristic this factor, so as to when query and title similarities are calculated, different grain size feature plays different effects according to the importance of oneself, so that Similarity Measure precision is higher, realize the optimization to existing voice similarity model, and can cause that Search Results are accurate, the demand of user can more be met.
Description
Technical field
The present invention relates to technical field of information processing, more particularly to a kind of semantic similarity acquisition side based on artificial intelligence
Method and device.
Background technology
Artificial intelligence (Artificial Intelligence), english abbreviation is AI.It is study, be developed for simulation,
Extend and extend a new technological sciences of intelligent theory, method, technology and the application system of people.Artificial intelligence is to calculate
One branch of machine science, it attempts to understand essence of intelligence, and produce it is a kind of it is new can be in the similar mode of human intelligence
The intelligence machine made a response, the research in the field includes robot, speech recognition, image recognition, natural language processing and specially
Family's system etc..
The search behavior of user is analyzed based on artificial intelligence, can learn that user's purpose in search is by searching
Hitch fruit can as early as possible get the content related to the search word that user is input into.
It is one of the key for realizing knowledge retrieval that semanteme according to search word retrieve, and Similarity Measure is then semantic
The basis of retrieval.Current semantic similarity model can calculate the search word (query) being input into when user searches for and be searched with candidate
Similarity between rope entry (title), after the similarity between obtaining query and title, search engine can be to obtaining
The similarity got is ranked up, and Search Results are returned according to ranking results.Fig. 1 is the structure of existing semantic similarity model
Schematic diagram.The semantic similarity model includes：Bottom be embedded (embedding) layer, conversion (BOW) layer, contrast (FC) layer with
And top layer marking (Score) layer.Wherein, the embedding layers of vectorization by all dictionary words is represented and constituted, when user is in search
When a sentence being input into after, the sentence can be mapped to a bivector by embedding layers, and each subvector is it
The corresponding term-embedding of term (term)；BOW layers represents the conversion made to bivector, and bivector is transformed into
One one-dimensional vector, the layer can also be replaced by convolution and pooling；FC layers is full UNICOM layer, and the FC layers to one-dimensional vector
Linear transformation is done, optionally non-linear turn can be added by the activation primitive to increase an activation primitive after Linear Transformation
Change；Score layers is measured for the similarity between the query and title to obtaining.For example, query is " Portugal of Brazil of Baidu
Language ", and title " Brazilian Portugal language ", after to query and title cutting words, can obtain the discrete word sequences of query and title,
The discrete word sequence of query includes：Baidu, Brazil, Portugal language, and title discrete word sequence includes：Brazil, Portugal language.By figure
When semantic similarity meter model shown in 1 calculates the similarity before query and title, by each word after query cutting words
As a granularity, a simple grain degree vector representation is then done to query using all of words of query, correspondingly, will
Then each word after title cutting words does a simple grain degree to title as a granularity using all of words of title
Vector representation.The Semantic Similarity Measurement of this simple grain degree, it is poor to get similarity precision, causes Search Results not enough to be managed
Think.
In order to improve search precision, as shown in Fig. 2 semantic scale model is improved, in the mistake of Similarity Measure
Cheng Zhong, after carrying out cutting word to query and title, feature extraction is carried out using participle language material, gets many of query and title
Binary feature (the query-basic- of base granularity feature (query-basic) query of individual grain size characteristic, such as query
Bigram), the binary feature (title-basic-bigram) of the base granularity feature (title-basic) of title, title.
It is similar between query and title is calculated although introducing many granularities to represent query and title as shown in Figure 2
Before degree, many grain size characteristics in semantic similarity model not to query and title do not make a distinction, straight in BOW layers of conversion
Connect and be added many grain size characteristics of query, obtain the representing multiple graininess of query, many grain size characteristics of title are added, obtain
The representing multiple graininess of title.
Directly be added for many granularities due to many grain size characteristics distinguish by existing voice similarity model, obtains
The representing multiple graininess of query and title so that the Search Results accuracy that search engine is obtained is poor.
The content of the invention
It is contemplated that at least solving one of technical problem in correlation technique to a certain extent.
Therefore, first purpose of the invention is to propose a kind of semantic similarity acquisition methods based on artificial intelligence,
To realize optimizing existing voice similarity model, for solving voice similarity model of the prior art due to not right
Many grain size characteristics distinguish and are directly added many granularities, obtain the representing multiple graininess of query and title so that search engine
The Search Results accuracy for obtaining is poor.
Second object of the present invention is to propose a kind of semantic similarity acquisition device based on artificial intelligence.
Third object of the present invention is to propose another semantic similarity acquisition device based on artificial intelligence.
Fourth object of the present invention is to propose a kind of non-transitorycomputer readable storage medium.
5th purpose of the invention is to propose a kind of computer program product.
It is that, up to above-mentioned purpose, first aspect present invention embodiment proposes a kind of semantic similarity based on artificial intelligence and obtains
Method is taken, including：
Obtain the grain size characteristic of search word and search entry；
Each grain size characteristic based on the search word carries out Similarity Measure with the grain size characteristic of the search entry, obtains
To the corresponding weight of each grain size characteristic；
The search word and the search entry are weighted using each grain size characteristic corresponding weight, are obtained
The granularity vector of search entry described in the granularity vector sum of the search word；
The granularity vector of search entry described in the granularity vector sum based on the search word, calculate the search word with it is described
Similarity between search entry.
The semantic similarity acquisition methods based on artificial intelligence of the embodiment of the present invention, by obtaining query and title
Grain size characteristic after, obtain the weight of each grain size characteristic, can embody varigrained feature by the weight is had
Significance level, then when representing multiple graininess is carried out to query and title, add each grain size characteristic weight this
Factor, so as to when query and title similarities are calculated, different grain size feature plays different works according to the importance of oneself
With so that Similarity Measure precision is higher, realizes the optimization to existing voice similarity model, and can cause Search Results
Precisely, the demand of user can more be met.
It is that, up to above-mentioned purpose, second aspect present invention embodiment proposes a kind of semantic similarity based on artificial intelligence and obtains
Device is taken, including：
Feature acquisition module, the grain size characteristic for obtaining search word and search entry；
Weight computation module, for the grain size characteristic of each grain size characteristic based on the search word and the search entry
Similarity Measure is carried out, the corresponding weight of each grain size characteristic is obtained；
Vectorial acquisition module, for utilizing the corresponding weight of each grain size characteristic to the search word and the search entry
It is weighted, obtains the granularity vector of search entry described in the granularity vector sum of the search word；
Similarity calculation module, for the granularity vector of search entry described in the granularity vector sum based on the search word,
Calculate the similarity between the search word and the search entry.
The semantic similarity acquisition device based on artificial intelligence of the embodiment of the present invention, by obtaining query and title
Grain size characteristic after, obtain the weight of each grain size characteristic, can embody varigrained feature by the weight is had
Significance level, then when representing multiple graininess is carried out to query and title, add each grain size characteristic weight this
Factor, so as to when query and title similarities are calculated, different grain size feature plays different works according to the importance of oneself
With so that Similarity Measure precision is higher, realizes the optimization to existing voice similarity model, and can cause Search Results
Precisely, the demand of user can more be met.
It is that, up to above-mentioned purpose, third aspect present invention embodiment proposes another semantic similarity based on artificial intelligence
Acquisition device, including：Processor；Memory for storing the processor-executable instruction；Wherein, the processor is matched somebody with somebody
It is set to：Obtain the grain size characteristic of search word and search entry；Each grain size characteristic and the searching bar based on the search word
Purpose grain size characteristic carries out Similarity Measure, obtains the corresponding weight of each grain size characteristic；It is corresponding using each grain size characteristic
Weight is weighted to the search word and the search entry, obtains being searched for described in the granularity vector sum of the search word
The granularity vector of entry；The granularity vector of search entry described in the granularity vector sum based on the search word, calculates the search
Similarity between word and the search entry.
To achieve these goals, fourth aspect present invention embodiment proposes a kind of non-transitory computer-readable storage
Medium, when the instruction in the storage medium is performed by the processor of server end so that server end is able to carry out one
The semantic similarity acquisition methods based on artificial intelligence are planted, methods described includes：Obtain search word special with the granularity of search entry
Levy；Each grain size characteristic based on the search word carries out Similarity Measure with the grain size characteristic of the search entry, obtains every
The corresponding weight of individual grain size characteristic；The search word and the search entry are carried out using each grain size characteristic corresponding weight
Weighted calculation, obtains the granularity vector of search entry described in the granularity vector sum of the search word；Grain based on the search word
The granularity vector of search entry described in degree vector sum, calculates the similarity between the search word and the search entry.
To achieve these goals, fifth aspect present invention embodiment proposes a kind of computer program product, when described
When instruction processing unit in computer program product is performed, a kind of semantic similarity acquisition methods based on artificial intelligence are performed,
Methods described includes：Obtain the grain size characteristic of search word and search entry；Each grain size characteristic and institute based on the search word
The grain size characteristic for stating search entry carries out Similarity Measure, obtains the corresponding weight of each grain size characteristic；It is special using each granularity
Levy corresponding weight to be weighted the search word and the search entry, obtain the granularity vector sum of the search word
The granularity vector of the search entry；The granularity vector of search entry described in the granularity vector sum based on the search word, calculates
Similarity between the search word and the search entry.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by practice of the invention.
Brief description of the drawings
The above-mentioned and/or additional aspect of the present invention and advantage will become from the following description of the accompanying drawings of embodiments
Substantially and be readily appreciated that, wherein：
Fig. 1 is the structural representation of the semantic similarity model of existing simple grain degree；
Fig. 2 is the structural representation of existing many granularity semantic similarity models；
Fig. 3 is shown by a kind of flow of semantic similarity acquisition methods based on artificial intelligence that the embodiment of the present invention is provided
It is intended to；
Fig. 4 is by a kind of structural representation of the embodiment of the present invention is provided semantic similarity model；
The flow of semantic similarity acquisition methods of the another kind based on artificial intelligence that Fig. 5 is provided by the embodiment of the present invention
Schematic diagram；
Fig. 6 is by the structural representation of the embodiment of the present invention is provided alternative semantic similarity model；
Fig. 7 is shown by a kind of structure of semantic similarity acquisition device based on artificial intelligence that the embodiment of the present invention is provided
It is intended to；
Fig. 8 is a kind of structural representation of vectorial acquisition module provided in an embodiment of the present invention；
Fig. 9 is a kind of structural representation of similarity calculation module provided in an embodiment of the present invention.
Specific embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from start to finish
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
It is exemplary to scheme the embodiment of description, it is intended to for explaining the present invention, and be not considered as limiting the invention.
Below with reference to the accompanying drawings the semantic similarity acquisition methods and device based on artificial intelligence of the embodiment of the present invention described.
Fig. 3 is shown by a kind of flow of semantic similarity acquisition methods based on artificial intelligence that the embodiment of the present invention is provided
It is intended to.The semantic similarity acquisition methods for being based on artificial intelligence are comprised the following steps：
S101, the grain size characteristic for obtaining search word and search entry.
Specifically, cutting word is carried out to search word (query) and search entry (title), obtains dividing for query and title
Word material, feature extraction is carried out using neutral net to participle language material, obtains the grain size characteristic of query and the granularity spy of title
Levy.
Before feature extraction is carried out to the participle language material of query and title by neutral net, it is necessary first to using instruction
Practice language material neutral net is trained, cutting word is carried out to training corpus before training, then using the morphology after cutting word into
The participle language material of many granularities, is trained based on participle language material to neutral net, and one is can be obtained by after the completion of training surely
Fixed convergent neutral net.
After query the and title cutting words of input, it is possible to which the corresponding participle language materials of query and title are input into god
Through carrying out feature extraction in network, the grain size characteristic of search word and search entry is obtained.For example, query is " Portugal of Brazil of Baidu
Language ", and title " Brazilian Portugal language ", after to query and title cutting words, can obtain the discrete word sequences of query and title,
The discrete word sequence of query includes：Baidu, Brazil, Portugal language, and title discrete word sequence includes：Brazil, Portugal language.Obtaining
To after query and title discrete word sequence, the word after participle can be combined, obtain the grain of query and title
Degree feature.For example, the grain size characteristic of query can include：The base granularity feature (query-basic, abbreviation qb) of query,
The binary grain size characteristic of the base granularity of the phrase grain size characteristic (query-phrase, abbreviation qp) and query of query
(query-basic-bigram, abbreviation abbreviation qbb) etc., correspondingly, the grain size characteristic of title can include：The base of title
Plinth grain size characteristic (title-basic, abbreviation tb), the phrase grain size characteristic (title-phrase becomes tp) of title and
Binary feature (title-basic-bigram, abbreviation tbb) of the base granularity of title etc..
The grain size characteristic of S102, each grain size characteristic based on search word and search entry carries out Similarity Measure, obtains
The corresponding weight of each grain size characteristic.
Specifically, each grain size characteristic of query is carried out into similar meter to the grain size characteristic of the same type of title respectively
Calculate, obtain the corresponding weight of each grain size characteristic.Preferably, by each grain size characteristic of query respectively with the same type of title
Grain size characteristic carry out cosine similarity calculating, obtain the corresponding weight of each grain size characteristic.For example, by query-basic with
Title-basic carries out cosine similarity calculating, obtains the weight of the basic grain size characteristics, by query-phrase with
Title-phrase carries out cosine similarity calculating, obtains the weight of the phrase grain size characteristics, and the weight can be embodied not
The significance level that the feature of one-size has.
S103, search word and search entry are weighted using each grain size characteristic corresponding weight, are searched
The granularity vector of the granularity vector sum search entry of rope word.
Specifically, after the corresponding weight of each grain size characteristic is got, by each grain size characteristic of query and the granularity
Weight corresponding to feature does product, and all grain size characteristics of query and corresponding weight are done into the value that product obtains is added,
Can be obtained by the granularity vector of the query.
Correspondingly, the weight corresponding to each grain size characteristic of title and the grain size characteristic is done into product, by title institutes
There are grain size characteristic and corresponding weight to do the value that product obtains to be added, it is possible to obtain the granularity vector representation of the title.
In the present embodiment, when the granularity vector of query and title is obtained, due to adding the power of each grain size characteristic
Heavy this factor, can represent the granularity vector of the granularity vector sum title of query according to the significance level of granularity so that
Granularity vector representation is more preferably accurate.
, it is necessary to the corresponding weight of each grain size characteristic to being calculated by cosine similarity is returned in practical application
One change is processed, and obtains the normalized weight of each grain size characteristic.Preferably, it is possible to use regression function (softmax) is to each
The weight of grain size characteristic is normalized.
The granularity vector of S104, the granularity vector sum search entry based on search word, calculate search word and search entry it
Between similarity.
After the granularity of granularity vector sum title of query vector is got, it is possible to the granularity based on the query
The granularity vector of vector sum title, calculates the similarity between the query and title.
A kind of structural representation of Fig. 4 its voice scale model provided for the present embodiment.As shown in figure 4, the semantic phase
Include like model：Granularity weighting layer, equivalent beds and scoring layer, the granularity to query and title is realized in the granularity weighting layer
Vector representation.Three cosine similarity computing modules, a normalization module and weighting mould are provided with the granularity weighting layer
Block, wherein, qb, respectively by a cosine similarity computing module, obtains basic granularities with tb, qp and tp and qbb and tbb
The weight of the digram features of the corresponding weight of feature, the corresponding weight of phrase grain size characteristics and basic granularities.Obtaining
To after weight, it is normalized by the weight for normalizing each grain size characteristic of the module to getting, then in weighting mould
Grain size characteristic is carried out the weighting that is multiplied by block with corresponding normalized weight, obtains granularity vector (q_1) and title of query
Granularity vector (t_1).The semantic scale model can perform the voice phase knowledge and magnanimity based on artificial intelligence of the present embodiment offer
Acquisition methods.
Specifically, the granularity vector of the granularity vector sum title of query is got in granularity weighting layer, then by query
Granularity vector sum title be input to equivalent beds i.e. FC layers of granularity vector, carry out linear transformation or non-by the FC layers
Linear transformation, obtains the similarity between query and title, then by score layers in semantic scale model to query
Similarity with title is scored, and after the completion of scoring, is ranked up according to scoring, and ranking results are fed back into user.
The semantic similarity acquisition methods based on artificial intelligence of the embodiment of the present invention, by obtaining query and title
Many grain size characteristics after, obtain the weight of each grain size characteristic, can embody varigrained feature by the weight is had
Some significance levels, then when representing multiple graininess is carried out to query and title, add each grain size characteristic weight this
One factor, so as to when query and title similarities are calculated, different grain size feature plays different according to the importance of oneself
Effect so that Similarity Measure precision is higher, realizes the optimization to existing voice similarity model, and can cause that search is tied
Fruit precisely, can more meet the demand of user.
The flow of semantic similarity acquisition methods of the another kind based on artificial intelligence that Fig. 5 is provided by the embodiment of the present invention
Schematic diagram.The semantic similarity acquisition methods for being based on artificial intelligence are comprised the following steps：
S201, the grain size characteristic for obtaining search word and search entry.
The grain size characteristic of S202, each grain size characteristic based on search word and search entry carries out Similarity Measure, obtains
The corresponding weight of each grain size characteristic.
S203, search word and search entry are weighted using each grain size characteristic corresponding weight, are searched
The granularity vector of the granularity vector sum search entry of rope word.
Introduction on S201~S203, can participate in the record of related content in above-described embodiment, and this is repeated no more.
The granularity vector of S204, each grain size characteristic to search word and search entry carries out Similarity Measure and obtains first
Weight, and each grain size characteristic to search entry carries out Similarity Measure and obtains the second power with the granularity vector of search entry
Weight.
S205, each grain size characteristic of search word and the first weight are weighted, and by search entry each
Grain size characteristic is weighted with the second weight.
Specifically, Similarity Measure is carried out to each grain size characteristic of query and the granularity vector of title and obtains the first power
Weight, and each grain size characteristic of title is carried out being calculated the second weight with the granularity vector similarity of query.By query
Each grain size characteristic be weighted with the first weight, will query each grain size characteristic and the first weight product, will
All product additions.Correspondingly, each grain size characteristic of title and the first weight are weighted, will title it is every
Individual grain size characteristic and the second weight product, by all product additions.
S206, using weighted calculation result update search word granularity vector sum search entry granularity vector.
After the result of weighted calculation of query and title is got, updated using the result of the weighted calculation of query
It is the granularity vector of query, the granularity vector of title is updated using the result of the weighted calculation of title.
S207, the similarity to the granularity vector of the granularity vector sum search entry of search word are calculated, and are searched for
Similarity between word and search entry.
After the granularity of granularity vector sum title of query vector is got, it is possible to the granularity based on the query
The granularity vector of vector sum title, calculates the similarity between the query and title.
The structural representation of Fig. 6 its alternative voice scale model for the present embodiment is provided.As shown in fig. 6, the semanteme
Scale model includes：First granularity weighting layer and the second granularity weighting layer, equivalent beds and scoring layer, in first granularity weighting
Layer and the second granularity weighting layer realize the granularity vector representation to query and title.It is provided with the first granularity weighting layer
Two the first cosine similarity computing modules, first normalization module and first weighting blocks, wherein, qb and tb and
Qbb and tbb by a first cosine similarity computing module, obtains the corresponding weight of basic grain size characteristics, basic respectively
The corresponding weight of digram features of granularity.After the corresponding weight of each grain size characteristic is got, by the first normalization module
Weight to each grain size characteristic is normalized, then in the first weighting block by grain size characteristic and corresponding normalization
Weight carries out product addition, obtains granularity vector (t_1) of a granularity of query vectorial (q_1) and title.
Further, the second granularity weighting layer includes：Four the second cosine similarity computing modules, two the second normalization
Module and two the second weighting blocks.
For query, by the digram features of basic grain size characteristics and basic granularities respectively with the first granularity weighting layer
The granularity i.e. t_1 of vector of the title for obtaining, carries out Similarity Measure and obtains by two the second cosine similarity computing modules respectively
To the first weight, the first weight is normalized by the second normalization module then, then in the second weighting block
By the qb and qbb of query respectively with corresponding normalization after the first weight carry out product addition, obtain a grain of query
Degree is vectorial (q_2).
Correspondingly, for title, by the digram features of basic grain size characteristics and basic granularities respectively with the first granularity
The granularity i.e. q_1 of vector of the query that weighting layer is obtained, carries out similarity by two the second cosine similarity computing modules respectively
The second weight is calculated, the second weight is normalized by the second normalization module then, then added second
Power module by the tb and tbb of title respectively with normalization after the second weight carry out product addition, obtain a granularity of title
Vectorial (t_2).
Equivalent beds i.e. FC layers that q_2 and t_2 are input to is being got, linear transformation or non-linear is being carried out by the FC layers
Conversion, obtain the similarity between query and title, then by score layer the in semantic scale model to query with
The similarity of title is scored, and after the completion of scoring, is ranked up according to scoring, and ranking results are fed back into user.
Further, in the present embodiment, operation can be iterated in the second granularity weighting block, i.e., to being obtained in S206
Query granularity vector sum title granularities vector i.e. using weighted calculation result update after search word granularity vector sum
The granularity vector of search entry is iterated calculating, and according to default iterations, iteration performs special to each granularity of query
Levy carries out Similarity Measure and obtains the first weight with the granularity vector of title, and each grain size characteristic to search entry with
The granularity vector of title carries out Similarity Measure and obtains the second weight, and by each grain size characteristic and the first weight of query
It is weighted, and each grain size characteristic of title and the second weight is weighted, until iterations completion is
Only.
For example, iterations is set to 2, will q_2 and t_2 as new q_1 and t_1, again by qb, qbb respectively with
Q_1 carries out cosine similarity calculating, and tb, tbb and t_1 are carried out into cosine similarity calculating and subsequent operation, obtains q_3
And t_3, an iteration process is now completed, q_3 and t_3 as new q_1 and t_1 again enters qb, qbb with q_1 respectively
Row cosine similarity is calculated, and tb, tbb and t_1 are carried out into cosine similarity calculating and subsequent operation, obtains q_4 and t_
4, now complete second iterative process.
The similarity meter between query and title after further q_4 and t_4 is performed as new q_1 and t_1
Calculate.
The semantic similarity acquisition methods based on artificial intelligence of the embodiment of the present invention, by obtaining query and title
Many grain size characteristics after, obtain the weight of each grain size characteristic, can embody varigrained feature by the weight is had
Some significance levels, then when representing multiple graininess is carried out to query and title, add each grain size characteristic weight this
One factor, so as to when query and title similarities are calculated, different grain size feature plays different according to the importance of oneself
Effect so that Similarity Measure precision is higher, realizes the optimization to existing voice similarity model, and can cause that search is tied
Fruit precisely, can more meet the demand of user.
In the present embodiment, multi-hop granularity weighting is carried out using the weight of grain size characteristic, further, improve similarity meter
The accuracy of calculation, so that Search Results more meet the demand of user.
Fig. 7 is a kind of structural representation of semantic similarity acquisition device based on artificial intelligence provided in an embodiment of the present invention
Figure.The semantic similarity acquisition device for being based on artificial intelligence includes：Feature acquisition module 11, weight computation module 12, vector
Acquisition module 13 and similarity calculation module 14.
Wherein, feature acquisition module 11, the grain size characteristic for obtaining search word and search entry.
Feature acquisition module 11, specifically for：
Cutting word is carried out to the search word and the search entry, the participle of the search word and the search entry is obtained
Language material.
Feature extraction is carried out to the participle language material using neutral net, the grain size characteristic of the search word and described is obtained
The grain size characteristic of search entry.
Weight computation module 12, it is special with the granularity of the search entry for each grain size characteristic based on the search word
Levying carries out Similarity Measure, obtains the corresponding weight of each grain size characteristic；
Weight computation module 12, specifically for by each grain size characteristic of the search word respectively with the search entry
The grain size characteristic of same type carries out similar calculating, obtains the corresponding weight of each grain size characteristic.
Further, weight computation module 12, specifically for：
Each grain size characteristic of the search word is carried out with the grain size characteristic of the same type of the search entry respectively remaining
String Similarity Measure, obtains the corresponding weight of each grain size characteristic.
Vectorial acquisition module 13, for utilizing the corresponding weight of each grain size characteristic to the search word and the searching bar
Mesh is weighted, and obtains the granularity vector of search entry described in the granularity vector sum of the search word.
Fig. 8 is a kind of optional structural representation of vectorial acquisition module in the present embodiment.The vectorial acquisition module 13 includes：
Normalization unit 131 and vectorial acquiring unit 132.
Normalization unit 131, for being normalized to the corresponding weight of each grain size characteristic, obtains each granularity
Corresponding normalized weight.
Vectorial acquiring unit 132, for for the search word and the search entry, by each grain size characteristic with it is corresponding
The normalized weight product addition, obtain search entry described in the granularity vector sum of the search word granularity vector.
Similarity calculation module 14, for search entry described in the granularity vector sum based on the search word granularity to
Amount, calculates the similarity between the search word and the search entry.
Fig. 9 is a kind of optional structural representation of similarity calculation module in the present embodiment.The similarity calculation module 14
Including：Weight unit 141, weight calculation unit 142, updating block 143, similarity calculated 144 and iteration unit
145。
Weight unit 141, enters for each grain size characteristic to the search word with the granularity vector of the search entry
Row Similarity Measure obtains the first weight, and each grain size characteristic to the search entry and the search entry granularity
Vector carries out Similarity Measure and obtains the second weight.
Weight calculation unit 142, for each grain size characteristic of the search word to be weighted with first weight
Calculate, and each grain size characteristic of the search entry is weighted with second weight.
Updating block 143, search described in the granularity vector sum of the search word is updated for the result using weighted calculation
The granularity vector of entry.
Similarity calculated 144, for the granularity vector of search entry described in the granularity vector sum to the search word
Similarity calculated, obtain the similarity between the search word and the search entry.
Further, iteration unit 145, for by the weighted calculation renewal after the search word granularity to
The granularity vector of amount and the search entry, described each grain to the search word is performed according to default iterations iteration
Degree feature carries out Similarity Measure and obtains the first weight with the granularity vector of the search entry, and to the search entry
Each grain size characteristic and the granularity vector of the search entry carry out Similarity Measure and obtain the second weight, and by the search
Each grain size characteristic of word is weighted with first weight, and by each grain size characteristic of the search entry and institute
State the second weight to be weighted, untill the iterations is completed.
Further, weight calculation unit 142, specifically for：
First weight and second weight are normalized；
Each grain size characteristic of the search word is weighted with first weight after normalization；
Each grain size characteristic of the search entry is weighted with second weight after normalization.
The semantic similarity acquisition methods based on artificial intelligence of the embodiment of the present invention, by obtaining query and title
Many grain size characteristics after, obtain the weight of each grain size characteristic, can embody varigrained feature by the weight is had
Some significance levels, then when representing multiple graininess is carried out to query and title, add each grain size characteristic weight this
One factor, so as to when query and title similarities are calculated, different grain size feature plays different according to the importance of oneself
Effect so that Similarity Measure precision is higher, realizes the optimization to existing voice similarity model, and can cause that search is tied
Fruit precisely, can more meet the demand of user.
In the present embodiment, multi-hop granularity weighting is carried out using the weight of grain size characteristic, further, improve similarity meter
The accuracy of calculation, so that Search Results more meet the demand of user.
In order to realize above-described embodiment, the present invention also proposes that another semantic similarity based on artificial intelligence obtains dress
Put, including：Processor, and for storing the memory of the processor-executable instruction.
Wherein, processor is configured as：Obtain the grain size characteristic of search word and search entry；Based on the every of the search word
Individual grain size characteristic carries out Similarity Measure with the grain size characteristic of the search entry, obtains the corresponding weight of each grain size characteristic；
The search word and the search entry are weighted using each grain size characteristic corresponding weight, obtain the search
The granularity vector of search entry described in the granularity vector sum of word；Search entry described in granularity vector sum based on the search word
Granularity vector, calculates the similarity between the search word and the search entry.
In order to realize above-described embodiment, the present invention also proposes a kind of non-transitorycomputer readable storage medium, when described
When instruction in storage medium is performed by the processor of server end so that server end is able to carry out a kind of based on artificial intelligence
The semantic similarity acquisition methods of energy, methods described includes：Obtain the grain size characteristic of search word and search entry；Searched based on described
Each grain size characteristic of rope word carries out Similarity Measure with the grain size characteristic of the search entry, obtains each grain size characteristic correspondence
Weight；The search word and the search entry are weighted using each grain size characteristic corresponding weight, are obtained
The granularity vector of search entry described in the granularity vector sum of the search word；Searched described in granularity vector sum based on the search word
Rope bar purpose granularity vector, calculates the similarity between the search word and the search entry.
In order to realize above-described embodiment, the present invention also proposes a kind of computer program product, when the computer program is produced
When instruction processing unit in product is performed, a kind of semantic similarity acquisition methods based on artificial intelligence are performed, methods described includes：
Obtain the grain size characteristic of search word and search entry；The grain of each grain size characteristic based on the search word and the search entry
Degree feature carries out Similarity Measure, obtains the corresponding weight of each grain size characteristic；Using the corresponding weight pair of each grain size characteristic
The search word and the search entry are weighted, and obtain search entry described in the granularity vector sum of the search word
Granularity vector；The granularity vector of search entry described in the granularity vector sum based on the search word, calculates the search word and institute
State the similarity between search entry.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means to combine specific features, structure, material or spy that the embodiment or example are described
Point is contained at least one embodiment of the invention or example.In this manual, to the schematic representation of above-mentioned term not
Identical embodiment or example must be directed to.And, the specific features of description, structure, material or feature can be with office
Combined in an appropriate manner in one or more embodiments or example.Additionally, in the case of not conflicting, the skill of this area
Art personnel can be tied the feature of the different embodiments or example described in this specification and different embodiments or example
Close and combine.
Additionally, term " first ", " second " are only used for describing purpose, and it is not intended that indicating or implying relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can express or
Implicitly include at least one this feature.In the description of the invention, " multiple " is meant that at least two, such as two, three
It is individual etc., unless otherwise expressly limited specifically.
Any process described otherwise above or method description in flow chart or herein is construed as, and expression includes
It is one or more for realizing custom logic function or process the step of the module of code of executable instruction, fragment or portion
Point, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discussion suitable
Sequence, including function involved by basis by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Represent in flow charts or logic and/or step described otherwise above herein, for example, being considered use
In the order list of the executable instruction for realizing logic function, in may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system, including the system of processor or other can be held from instruction
The system of row system, device or equipment instruction fetch and execute instruction) use, or with reference to these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicate, propagate or pass
The dress that defeated program is used for instruction execution system, device or equipment or with reference to these instruction execution systems, device or equipment
Put.The more specifically example (non-exhaustive list) of computer-readable medium includes following：With the electricity that one or more are connected up
Connecting portion (electronic installation), portable computer diskette box (magnetic device), random access memory (RAM), read-only storage
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device, and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can thereon print described program or other are suitable
Medium, because optical scanner for example can be carried out by paper or other media, then enters edlin, interpretation or if necessary with it
His suitable method is processed electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each several part of the invention can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In implementation method, the software that multiple steps or method can in memory and by suitable instruction execution system be performed with storage
Or firmware is realized.Such as, if being realized, with another embodiment, following skill well known in the art being used with hardware
Any one of art or their combination are realized：With the logic gates for realizing logic function to data-signal from
Scattered logic circuit, the application specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method is carried
The rapid hardware that can be by program to instruct correlation is completed, and described program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
Additionally, during each functional unit in each embodiment of the invention can be integrated in a processing module, it is also possible to
It is that unit is individually physically present, it is also possible to which two or more units are integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, it would however also be possible to employ the form of software function module is realized.The integrated module is such as
Fruit is to realize in the form of software function module and as independent production marketing or when using, it is also possible to which storage is in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..Although having been shown above and retouching
Embodiments of the invention are stated, it is to be understood that above-described embodiment is exemplary, it is impossible to be interpreted as to limit of the invention
System, one of ordinary skill in the art can be changed to above-described embodiment, change, replace and become within the scope of the invention
Type.
Claims (16)
1. a kind of semantic similarity acquisition methods based on artificial intelligence, it is characterised in that including：
Obtain the grain size characteristic of search word and search entry；
Each grain size characteristic based on the search word carries out Similarity Measure with the grain size characteristic of the search entry, obtains every
The corresponding weight of individual grain size characteristic；
The search word and the search entry are weighted using each grain size characteristic corresponding weight, obtain described
The granularity vector of search entry described in the granularity vector sum of search word；
The granularity vector of search entry described in the granularity vector sum based on the search word, calculates the search word and the search
Similarity between entry.
2. semantic similarity acquisition methods based on artificial intelligence according to claim 1, it is characterised in that the acquisition
The grain size characteristic of search word and search entry, including：
Cutting word is carried out to the search word and the search entry, the participle language of the search word and the search entry is obtained
Material；
Feature extraction is carried out to the participle language material using neutral net, the grain size characteristic of the search word and the search is obtained
The grain size characteristic of entry.
3. semantic similarity acquisition methods based on artificial intelligence according to claim 1, it is characterised in that described to be based on
Each grain size characteristic of the search word carries out Similarity Measure with the grain size characteristic of the search entry, obtains each granularity special
Corresponding weight is levied, including：
Each grain size characteristic of the search word is carried out into similar meter to the grain size characteristic of the same type of the search entry respectively
Calculate, obtain the corresponding weight of each grain size characteristic.
4. semantic similarity acquisition methods based on artificial intelligence according to claim 3, it is characterised in that described by institute
State each grain size characteristic of search word carries out similar being calculated often to the grain size characteristic of the same type of the search entry respectively
The corresponding weight of individual grain size characteristic, including：
Each grain size characteristic of the search word is carried out into cosine phase with the grain size characteristic of the same type of the search entry respectively
Calculated like degree, obtain the corresponding weight of each grain size characteristic.
5. semantic similarity acquisition methods based on artificial intelligence according to claim 4, it is characterised in that the utilization
The corresponding weight of each grain size characteristic is weighted to the search word and the search entry, obtains the search word
The granularity vector of search entry described in granularity vector sum, including：
The corresponding weight of each grain size characteristic is normalized, the corresponding normalized weight of each granularity is obtained；
For the search word and the search entry, by the product phase of each grain size characteristic and the corresponding normalized weight
Plus, obtain the granularity vector of search entry described in the granularity vector sum of the search word.
6. semantic similarity acquisition methods based on artificial intelligence according to claim 5, it is characterised in that described to be based on
The granularity vector of search entry described in the granularity vector sum of the search word, calculates between the search word and the search entry
Similarity, including：
Similarity Measure is carried out with the granularity vector of the search entry to each grain size characteristic of the search word and obtains first
Weight, and each grain size characteristic to the search entry carries out Similarity Measure and obtains with the granularity vector of the search entry
To the second weight；
Each grain size characteristic of the search word is weighted with first weight, and by the every of the search entry
Individual grain size characteristic is weighted with second weight；
The granularity vector of search entry described in the granularity vector sum of the search word is updated using the result of weighted calculation；
Similarity to the granularity vector of search entry described in the granularity vector sum of the search word is calculated, and obtains described searching
Similarity between rope word and the search entry.
7. semantic similarity acquisition methods based on artificial intelligence according to claim 6, it is characterised in that the utilization
Before the granularity vector of search entry described in the granularity vector sum of the result renewal search word of weighted calculation, also include：
To the granularity vector by search entry described in the granularity vector sum of the search word after weighted calculation renewal, press
According to default iterations iteration perform the granularity of described each grain size characteristic to the search word and the search entry to
Amount carries out Similarity Measure and obtains the first weight, and each grain size characteristic and the search entry to the search entry
Granularity vector carries out Similarity Measure and obtains the second weight, and each grain size characteristic of the search word and described first are weighed
It is weighted again, and each grain size characteristic of the search entry is weighted with second weight, until
Untill the iterations is completed.
8. voice similarity acquisition methods based on artificial intelligence according to claim 5 or 6, it is characterised in that described
Each grain size characteristic of the search word is weighted with first weight, and by each grain of the search entry
Degree feature is weighted with second weight, including：
First weight and second weight are normalized；
Each grain size characteristic of the search word is weighted with first weight after normalization；
Each grain size characteristic of the search entry is weighted with second weight after normalization.
9. a kind of semantic similarity acquisition device based on artificial intelligence, it is characterised in that including：
Feature acquisition module, the grain size characteristic for obtaining search word and search entry；
Weight computation module, is carried out for each grain size characteristic based on the search word with the grain size characteristic of the search entry
Similarity Measure, obtains the corresponding weight of each grain size characteristic；
Vectorial acquisition module, for being carried out to the search word and the search entry using the corresponding weight of each grain size characteristic
Weighted calculation, obtains the granularity vector of search entry described in the granularity vector sum of the search word；
Similarity calculation module, for the granularity vector of search entry described in the granularity vector sum based on the search word, calculates
Similarity between the search word and the search entry.
10. the semantic similarity acquisition device based on artificial intelligence according to claim 9, it is characterised in that the spy
Acquisition module is levied, specifically for：
Cutting word is carried out to the search word and the search entry, the participle language of the search word and the search entry is obtained
Material；
Feature extraction is carried out to the participle language material using neutral net, the grain size characteristic of the search word and the search is obtained
The grain size characteristic of entry.
The 11. semantic similarity acquisition device based on artificial intelligence according to claim 9, it is characterised in that the power
Re-computation module, specifically for：
Each grain size characteristic of the search word is carried out into similar meter to the grain size characteristic of the same type of the search entry respectively
Calculate, obtain the corresponding weight of each grain size characteristic.
The 12. semantic similarity acquisition device based on artificial intelligence according to claim 11, it is characterised in that the power
Re-computation module, specifically for：
Each grain size characteristic of the search word is carried out into cosine phase with the grain size characteristic of the same type of the search entry respectively
Calculated like degree, obtain the corresponding weight of each grain size characteristic.
The 13. semantic similarity acquisition device based on artificial intelligence according to claim 12, it is characterised in that it is described to
Amount acquisition module, including：
Normalization unit, for being normalized to the corresponding weight of each grain size characteristic, obtains each granularity corresponding
Normalized weight；
Vectorial acquiring unit, it is for for the search word and the search entry, each grain size characteristic is described with corresponding
The product addition of normalized weight, obtains the granularity vector of search entry described in the granularity vector sum of the search word.
The 14. semantic similarity acquisition device based on artificial intelligence according to claim 13, it is characterised in that the phase
Like degree computing module, including：
Weight unit, similarity is carried out for each grain size characteristic to the search word with the granularity vector of the search entry
It is calculated the first weight, and each grain size characteristic to the search entry is carried out with the granularity vector of the search entry
Similarity Measure obtains the second weight；
Weight calculation unit, for each grain size characteristic of the search word to be weighted with first weight, and
Each grain size characteristic of the search entry is weighted with second weight；
Updating block, the grain of search entry described in the granularity vector sum of the search word is updated for the result using weighted calculation
Degree vector；
Similarity calculated, for the similarity of the granularity vector of search entry described in the granularity vector sum to the search word
Calculated, obtained the similarity between the search word and the search entry.
The 15. semantic similarity acquisition device based on artificial intelligence according to claim 14, it is characterised in that the phase
Like degree computing module, also include：
Iteration unit, for by search entry described in the granularity vector sum of the search word after weighted calculation renewal
Granularity vector, according to default iterations iteration perform described each grain size characteristic to the search word and the search
The granularity vector of entry carries out Similarity Measure and obtains the first weight, and each grain size characteristic to the search entry and institute
State the granularity vector of search entry and carry out Similarity Measure and obtain the second weight, and by each grain size characteristic of the search word
It is weighted with first weight, and each grain size characteristic of the search entry is added with second weight
Power is calculated, untill the iterations is completed.
The 16. voice similarity acquisition device based on artificial intelligence according to claims 14 or 15, it is characterised in that institute
Weight calculation unit is stated, specifically for：
First weight and second weight are normalized；
Each grain size characteristic of the search word is weighted with first weight after normalization；
Each grain size characteristic of the search entry is weighted with second weight after normalization.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611042515.3A CN106776782B (en) | 2016-11-21 | 2016-11-21 | Semantic similarity obtaining method and device based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611042515.3A CN106776782B (en) | 2016-11-21 | 2016-11-21 | Semantic similarity obtaining method and device based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106776782A true CN106776782A (en) | 2017-05-31 |
| CN106776782B CN106776782B (en) | 2020-05-22 |
Family
ID=58974777
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611042515.3A Active CN106776782B (en) | 2016-11-21 | 2016-11-21 | Semantic similarity obtaining method and device based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106776782B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111078849A (en) * | 2019-12-02 | 2020-04-28 | 百度在线网络技术（北京）有限公司 | Method and apparatus for outputting information |
| CN111128376A (en) * | 2019-11-21 | 2020-05-08 | 泰康保险集团股份有限公司 | Method and device for recommending evaluation form |
| CN111259118A (en) * | 2020-05-06 | 2020-06-09 | 广东电网有限责任公司 | Text data retrieval method and device |
| CN111931034A (en) * | 2020-08-24 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Data searching method, device, equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011248596A (en) * | 2010-05-26 | 2011-12-08 | Hitachi Ltd | Searching system and searching method for picture-containing documents |
| CN103838789A (en) * | 2012-11-27 | 2014-06-04 | 大连灵动科技发展有限公司 | Text similarity computing method |
| CN104462060A (en) * | 2014-12-03 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Method and device for calculating text similarity and realizing search processing through computer |
| CN104899322A (en) * | 2015-06-18 | 2015-09-09 | 百度在线网络技术（北京）有限公司 | Search engine and implementation method thereof |
-
2016
- 2016-11-21 CN CN201611042515.3A patent/CN106776782B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011248596A (en) * | 2010-05-26 | 2011-12-08 | Hitachi Ltd | Searching system and searching method for picture-containing documents |
| CN103838789A (en) * | 2012-11-27 | 2014-06-04 | 大连灵动科技发展有限公司 | Text similarity computing method |
| CN104462060A (en) * | 2014-12-03 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Method and device for calculating text similarity and realizing search processing through computer |
| CN104899322A (en) * | 2015-06-18 | 2015-09-09 | 百度在线网络技术（北京）有限公司 | Search engine and implementation method thereof |
Non-Patent Citations (1)
| Title |
|---|
| 杨春龙等: "基于概念语义相似度计算模型的信息检索研究", 《计算机应用与软件》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111128376A (en) * | 2019-11-21 | 2020-05-08 | 泰康保险集团股份有限公司 | Method and device for recommending evaluation form |
| CN111078849A (en) * | 2019-12-02 | 2020-04-28 | 百度在线网络技术（北京）有限公司 | Method and apparatus for outputting information |
| CN111259118A (en) * | 2020-05-06 | 2020-06-09 | 广东电网有限责任公司 | Text data retrieval method and device |
| CN111931034A (en) * | 2020-08-24 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Data searching method, device, equipment and storage medium |
| CN111931034B (en) * | 2020-08-24 | 2024-01-26 | 腾讯科技（深圳）有限公司 | Data searching method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106776782B (en) | 2020-05-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111125331B (en) | Semantic recognition method, semantic recognition device, electronic equipment and computer readable storage medium | |
| CN112214610B (en) | Entity relationship joint extraction method based on span and knowledge enhancement | |
| CN110750640B (en) | Text data classification method and device based on neural network model and storage medium | |
| CN101566998B (en) | Chinese question-answering system based on neural network | |
| CN107102989A (en) | A kind of entity disambiguation method based on term vector, convolutional neural networks | |
| CN106970910A (en) | A kind of keyword extracting method and device based on graph model | |
| CN108509411A (en) | Semantic analysis and device | |
| CN109344236A (en) | One kind being based on the problem of various features similarity calculating method | |
| CN109376222A (en) | Question and answer matching degree calculation method, question and answer automatic matching method and device | |
| CN109670191A (en) | Calibration optimization method, device and the electronic equipment of machine translation | |
| CN106776782A (en) | Semantic similarity acquisition methods and device based on artificial intelligence | |
| US11734322B2 (en) | Enhanced intent matching using keyword-based word mover's distance | |
| CN106844341A (en) | News in brief extracting method and device based on artificial intelligence | |
| CN104484380A (en) | Personalized search method and personalized search device | |
| CN108959556A (en) | Entity answering method, device and terminal neural network based | |
| CN109597988A (en) | The former prediction technique of vocabulary justice, device and electronic equipment across language | |
| CN107122492A (en) | Lyric generation method and device based on picture content | |
| CN110674301A (en) | Emotional tendency prediction method, device and system and storage medium | |
| CN110929498A (en) | Short text similarity calculation method and device and readable storage medium | |
| CN108470025A (en) | Partial-Topic probability generates regularization own coding text and is embedded in representation method | |
| CN109766537A (en) | Study abroad document methodology of composition, device and electronic equipment | |
| CN107193941A (en) | Story generation method and device based on picture content | |
| CN110399477A (en) | A kind of literature summary extracting method, equipment and can storage medium | |
| CN115827988B (en) | Self-media content heat prediction method | |
| US10685281B2 (en) | Automated predictive modeling and framework |
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 |