CN107515857A - Semantic understanding method and system based on customization technical ability - Google Patents

Semantic understanding method and system based on customization technical ability Download PDF

Info

Publication number
CN107515857A
CN107515857A CN201710774597.9A CN201710774597A CN107515857A CN 107515857 A CN107515857 A CN 107515857A CN 201710774597 A CN201710774597 A CN 201710774597A CN 107515857 A CN107515857 A CN 107515857A
Authority
CN
China
Prior art keywords
intended
intention
customization
ability
template
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
Application number
CN201710774597.9A
Other languages
Chinese (zh)
Other versions
CN107515857B (en
Inventor
黄鑫
陈志刚
王智国
胡国平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
iFlytek Co Ltd
Original Assignee
iFlytek Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by iFlytek Co Ltd filed Critical iFlytek Co Ltd
Priority to CN201710774597.9A priority Critical patent/CN107515857B/en
Publication of CN107515857A publication Critical patent/CN107515857A/en
Application granted granted Critical
Publication of CN107515857B publication Critical patent/CN107515857B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/30Semantic analysis

Abstract

The invention discloses a kind of semantic understanding method and system based on customization technical ability, this method includes:Receive user's request data;The path of current customization technical ability is obtained, the semantic resources of current customization technical ability are obtained according to the path;Determine that the request data belongs to the saying template of current customization technical ability, one or more semantic grooves, each corresponding entity of semantic groove are included in the saying template;The semantic groove and correspondent entity in the request data are determined according to the saying template, obtains semantic understanding result.Utilize the present invention, it is possible to achieve to user's request data fast and accurately semantic understanding.

Description

Semantic understanding method and system based on customization technical ability
Technical field
The present invention relates to natural language processing field, and in particular to it is a kind of based on customization technical ability semantic understanding method and be System.
Background technology
With the development of mobile intelligent terminal and information network technique, people are handed in increasing scene using voice Mutually application, such as using intelligent sound box as entrance, interactive voice uses weather/stock/music/traffic/alarm clock/prompting etc. Using every kind of application includes one or more technical ability.In another example using vehicle-mounted vehicle device as entrance, interactive voice using LBS/ navigation/ The applications such as radio station/music or technical ability.Thus, also there are increasing dialog mode artificial intelligence open platform, such as hundred in industry The DuerOS of degree, the Alexa of Amazon etc., providing one for application developer or skill development person can be with development and application or technical ability System.Meanwhile a kind of abundant ecological purpose for sharing win-win each other is built for open platform, it is also typically provided technical ability business The market of this similar type in city so that the technical ability of developer's exploitation, it can be published in store, both can be directly for terminal Family selection uses, and can also be chosen for third party application developer in application, terminal user is available to by terminal applies Use.
In order to further meet the individual demand of different application, allow application developer using the technical ability when according to The demand of oneself is customized, and some technical ability furthermore provide customization function, such as, authority is customized for technical ability setting, Application developer can carry out personalized customization when using the technical ability in the customization extent of competence, obtain customizing technical ability. Certainly, the technical ability after customization can be not only applied in application program newly developed by application developer, can also customize this Technical ability is published in technical ability store as shared technical ability, selects to use for other developers or terminal user.
For the application in terms of natural language understanding, such as interactive voice application, user is generally used using client should Using client receives user's request data, carries out semantic understanding to user's request data using the technical ability in the application, returns Respond to user.
For customizing technical ability, how fast and accurately semantic understanding is carried out to user's request data, there is presently no effective Solution.
The content of the invention
The embodiment of the present invention provides a kind of semantic understanding method and system based on customization technical ability, and user is asked with realizing Data fast and accurately semantic understanding.
Therefore, the present invention provides following technical scheme:
A kind of semantic understanding method based on customization technical ability, including:
Receive user's request data;
The path of current customization technical ability is obtained, the semantic resources of current customization technical ability are obtained according to the path;
Determine that the request data belongs to the saying template of current customization technical ability, one or more is included in the saying template Individual semantic groove, each corresponding entity of semantic groove;
The semantic groove and correspondent entity in the request data are determined according to the saying template, obtains semantic understanding knot Fruit.
Preferably, it is described to determine that the request data belongs to the current saying template for customizing technical ability and included:
Determine that the request data belongs to the intention of current customization technical ability;
The request data each saying template corresponding with the intention is matched, it is corresponding to obtain the request data Saying template.
Preferably, methods described also includes:
Previously according to the saying template of the intention of current customization technical ability, exercise skill is intended to determine model;And utilize technical ability The intention probability structure for being intended to determine to export during model training is intended to order models;
It is described to determine that the request data belongs to the current intention for customizing technical ability and included:
It is intended to determine model using the technical ability, determines that the request data belongs to during customization technical ability is intended to what is be each intended to Probability;
The request data is belonged to during customization technical ability is intended to the probability being each intended to and inputs the intention order models, root Intention probability after being sorted according to the output of the intention order models;
It is intended to the intention as the request data corresponding to the most forward intention probability of selected and sorted.
Preferably, the intention of the current customization technical ability includes:Customization is intended to and is customized intention;The customization refers to The current stylish increased intention for being intended to and modifying to obtain to being customized technical ability intention of customization technical ability customization;The quilt The intention that the foundational skills that customization refers to use during current customization technical ability customization are not changed in being intended to;
Methods described also includes:
Previously according to the saying template that customization is intended to and the saying template for being customized intention, training customization is intended to determine model It is intended to determine model with being customized;And the meaning for being intended to determine model with export when being customized and being intended to determine model training using customization Figure probability structure is intended to order models;
It is described to determine that the request data belongs to the current intention for customizing technical ability and included:
It is intended to determine model and be customized to be intended to determine model using the customization, determines that the request data belongs to customization The probability that the probability that is each intended in intention and being customized each is intended in being intended to;
The request data is belonging respectively to customize the probability being each intended in intention and be customized in intention to be each intended to Probability input the intention order models, the intention probability after being sorted according to the output of the intention order models;
It is intended to the intention as the request data corresponding to the most forward intention probability of selected and sorted.
Preferably, methods described also includes:
Previously according to the saying template of the intention of current customization technical ability, exercise skill saying template determines model, and utilizes Technical ability saying template determines the saying template probabilities structure saying template order models exported during model training;
It is described to determine that the request data belongs to the current saying template for customizing technical ability and included:
Model is determined using the technical ability saying template, determines that the request data belongs to each in the intention for customizing technical ability The probability of saying template;
The probability that the request data is belonged to each saying template in the intention of customization technical ability inputs the saying template Order models, the saying template probabilities after being sorted according to the output of the saying template order models;
Saying template is as saying corresponding to the request data corresponding to the most forward saying template probabilities of selected and sorted Template.
Preferably, currently the intention of customization technical ability includes:Customization is intended to and is customized intention;The customization refers to current The stylish increased intention of customization technical ability customization and the intention for modifying to obtain to being customized technical ability intention;It is described to be customized The intention that the foundational skills for referring to use during current customization technical ability customization are not changed in being intended to;
Methods described also includes:
Previously according to the saying template that customization is intended to and the saying template for being customized intention, training customization is intended to saying template Determine that model determines model with intention saying template is customized;And determine model using customization intention saying template and be customized meaning Figure saying template determines the saying template probabilities structure saying template order models exported during model training;
It is described to determine that the request data belongs to the current saying template for customizing technical ability and included:
It is intended to saying template using the customization and determines that model determines model with intention saying template is customized, it is determined that described Request data belongs to the probability of each saying template and the probability for being customized each saying template in intention during customization is intended to;
The request data is belonging respectively to the probability of each saying template during customization is intended to and is customized each in intention The probability of saying template inputs the saying template order models, is sorted according to the output of the saying template order models Saying template probabilities afterwards;
Saying template is as the corresponding saying template corresponding to the most forward saying template probabilities of selected and sorted.
A kind of semantic understanding system based on customization technical ability, including:
Receiving module, for receiving user's request data;
Source obtaining module, for obtaining the path of current customization technical ability, current customization technical ability is obtained according to the path Semantic resources;
Saying template determining module, it is described for determining that the request data belongs to the saying template of current customization technical ability One or more semantic grooves, each corresponding entity of semantic groove are included in saying template;
Semantic understanding module, for determining the semantic groove in the request data according to the saying template and corresponding to real Body, obtain semantic understanding result.
Preferably, the saying template determining module includes:
It is intended to determining module, for determining that the request data belongs to the intention of current customization technical ability;
Template matches module, for the request data each saying template corresponding with the intention to be matched, obtain To saying template corresponding to the request data.
Preferably, the system also includes:
First model construction module, for the saying template of the intention previously according to current customization technical ability, exercise skill meaning Figure determines model;And the intention probability structure for being intended to determine to export during model training using technical ability is intended to order models;
The intention determining module includes:
First intention probability determining unit, for being intended to determine model using the technical ability, determine the request data category The probability being each intended in customization technical ability is intended to;
First sequencing unit, institute is inputted for the request data to be belonged to during customization technical ability is intended into the probability being each intended to Intention order models are stated, the intention probability after being sorted according to the output of the intention order models;
First choice unit, most forward for selected and sorted is intended to be intended to as the request data corresponding to probability It is intended to.
Preferably, the intention of the current customization technical ability includes:Customization is intended to and is customized intention;The customization refers to The current stylish increased intention for being intended to and modifying to obtain to being customized technical ability intention of customization technical ability customization;The quilt The intention that the foundational skills that customization refers to use during current customization technical ability customization are not changed in being intended to;
The system also includes:
Second model construction module, for the saying template being intended to previously according to customization and the saying mould for being customized intention Plate, training customization are intended to determine that model determines model with intention is customized;And it is intended to determine model using customization and is customized meaning The intention probability structure that figure determines to export during model training is intended to order models;
The intention determining module includes:
Second intention probability determining unit, for being intended to determine model and be customized to be intended to determine mould using the customization Type, determine that the request data belongs to during customization is intended to the probability being each intended to and is customized the probability being each intended in intention;
Second sequencing unit, for the request data to be belonging respectively to customize the probability being each intended in intention and be determined The probability that system is each intended in being intended to inputs the intention order models, is sorted according to the output of the intention order models Intention probability afterwards;
Second selecting unit, most forward for selected and sorted is intended to be intended to as the request data corresponding to probability It is intended to.
Preferably, the system also includes:
3rd model construction module, for the saying template of the intention previously according to current customization technical ability, exercise skill is said Method template determines model, and the saying template probabilities for determining to export during model training using technical ability saying template build saying template Order models;
The saying template determining module includes:
3rd probability determining unit, for determining model using the technical ability saying template, determine the request data category The probability of each saying template in the intention of customization technical ability;
3rd sequencing unit, for the request data to be belonged to the probability of each saying template in the intention for customizing technical ability The saying template order models are inputted, the saying template after being sorted according to the output of the saying template order models is general Rate;
3rd selecting unit, saying template corresponding to the saying template probabilities most forward for selected and sorted are used as described ask Seek saying template corresponding to data.
Preferably, currently the intention of customization technical ability includes:Customization is intended to and is customized intention;The customization refers to current The stylish increased intention of customization technical ability customization and the intention for modifying to obtain to being customized technical ability intention;It is described to be customized The intention that the foundational skills for referring to use during current customization technical ability customization are not changed in being intended to;
The system also includes:
4th model construction module, for the saying template being intended to previously according to customization and the saying mould for being customized intention Plate, training customization are intended to saying template and determine that model determines model with intention saying template is customized;And it is intended to say using customization Method template determines model and is customized to be intended to the saying template probabilities structure saying mould that saying template determines to export during model training Plate order models;
The saying template determining module includes:
4th probability determining unit, for determining model using the customization intention saying template and being customized intention saying Template determines model, determines that the request data belongs to the probability of each saying template during customization is intended to and is customized every in intention The probability of individual saying template;
4th sequencing unit, for by the request data be belonging respectively to during customization is intended to the probability of each saying template and The probability for being customized each saying template in being intended to inputs the saying template order models, according to saying template sequence mould The output of type sorted after saying template probabilities;
4th selecting unit, saying template is as described right corresponding to the saying template probabilities most forward for selected and sorted The saying template answered.
Semantic understanding method and system provided in an embodiment of the present invention based on customization technical ability, are receiving user's number of request According to rear, the path of the current customization technical ability of acquisition, and all semantic resources for customizing technical ability are obtained according to the path, determine institute The saying template that request data belongs to current customization technical ability is stated, the semanteme in the request data is determined according to the saying template Groove and correspondent entity, obtain semantic understanding result.Utilize the present invention, it is possible to achieve to user's request data fast and accurately language Reason and good sense solution.
Brief description of the drawings
, below will be to institute in embodiment in order to illustrate more clearly of the embodiment of the present application or technical scheme of the prior art The accompanying drawing needed to use is briefly described, it should be apparent that, drawings in the following description are only one described in the present invention A little embodiments, for those of ordinary skill in the art, other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 is the flow chart of semantic understanding method of the embodiment of the present invention based on customization technical ability;
Fig. 2 is a kind of flow chart that saying template corresponding to request data is determined in the embodiment of the present invention;
Fig. 3 is another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention;
Fig. 4 is another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention;
Fig. 5 is another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention;
Fig. 6 is the structural representation of semantic understanding system of the embodiment of the present invention based on customization technical ability.
Embodiment
In order that those skilled in the art more fully understand the scheme of the embodiment of the present invention, below in conjunction with the accompanying drawings and implement Mode is described in further detail to the embodiment of the present invention.
As shown in figure 1, the flow chart for the semantic understanding method based on customization technical ability that is the embodiment of the present invention, including following step Suddenly:
Step 101, user's request data is received.
User's request data is used to express the current request of user, be specifically as follows speech data or text data, Or other multimedia request data, such as view data.As user says request speech data " I wants to eat Spicy diced chicken with peanuts ";When So, if user's request data is speech data, it is also necessary to the speech data is identified as into text data, if picture number According to, it is necessary to which view data is identified as into text data by image-recognizing method.
Step 102, the path of current customization technical ability is obtained, obtaining current the semantic of customization technical ability according to the path provides Source.
The customization technical ability, refer to application developer in other skills bases, the skill developed according to itself application demand Energy.It generally, can be selected according to application demand corresponding comprising a large amount of shared technical ability classified, application developer in technical ability store After shared technical ability under classification, it is customized with reference to self-demand, intends exploitation such as application developer and listen this application of music, it is clear Look at behind technical ability store, find music walkman technical ability, the technical ability, which includes, listens music services, and application developer is on the basis of the technical ability On, technical ability customization is carried out with reference to self-demand, such as friend recommendation song is added in music walkman technical ability, broadcasts one section of music Allow the social functions such as good friend's game of guessing the titles of songs name.
The semantic resources of one technical ability, i.e. technical ability carry out the resource used during semantic understanding, saying template, language such as technical ability Adopted groove, entity etc., it is typically stored on a server.Because the technical ability number of an open platform is more, it will usually use Distributed storage mode, i.e., the semantic resources of different technical ability are potentially stored on different servers;Thus for multiple customization Technical ability, that is, customize technical ability, its semantic resources is typically stored on multiple servers.Therefore, can be with when carrying out semantic understanding All semantic resources of current customization technical ability are first obtained from different server, and these semantic resources are put into same server On, it is convenient when being performed in order to program that semantic resources are loaded into internal memory.
Service where have registered the semantic resources of the technical ability to resource management server in exploitation due to each technical ability The address of device, customization technical ability are no exception.Therefore, can be according to current customization when obtaining the semantic resources of current customization technical ability The path of technical ability, server address where the semantic resources of each technical ability on the path is sequentially found, according to the server Location can obtain all semantic resources of the customization technical ability on the path, and the semantic resources are put into same server On.
Further, in order to subsequently easily customize technical ability using current, the numbering of current customization technical ability can also be preserved And its server address where corresponding semantic resources, can be according to the server of preservation when reusing the customization technical ability next time Respective server acquisition is directly arrived in address.
Step 103, determine that the request data belongs to the saying template of current customization technical ability, included in the saying template One or more semantic grooves, each corresponding entity of semantic groove.
Because each technical ability can include multiple intentions, and each it is intended to include a variety of saying templates.Therefore, it is determined that institute When stating request data and belonging to the saying template of current customization technical ability, it can first determine that the request data belongs to current customization technical ability Intention, then the request data each saying template corresponding with the intention is matched again, obtains the number of request According to corresponding saying template;It can also directly determine the request data belongs to which saying template of current customization technical ability.
Above-mentioned several different modes will be described in detail later.
Step 104, the semantic groove and correspondent entity in the request data are determined according to the saying template, obtains semanteme Understand result.
Specifically, the method that accurate matching or fuzzy matching can be used, determine semantic groove in the request data and Correspondent entity.For example user's request data is " my main points chicken cubes in chilly sauce ", the intention that is intended to order dishes of request data, according to Each saying template for intention of ordering dishes, it may be determined that the saying template that active user's request data belongs to current customization technical ability is " I Main points dish " wherein dish are semantic groove, and user's request data " my main points chicken cubes in chilly sauce " and the saying template are carried out Matching, it is " chicken cubes in chilly sauce " to obtain semantic groove dish correspondent entities.
If it should be noted that it is determined that during saying template belonging to user's request data, by user's request data with Saying template is matched, then directly can obtain the semantic groove in user's request data according to matching result and correspond to real Body, it is not necessary to matched again.
As shown in Fig. 2 be a kind of flow chart that saying template corresponding to request data is determined in the embodiment of the present invention, including Following steps:
Step 201, it is intended to previously according to the saying template that customization is intended to and the saying template for being customized intention, training customization Determine model and be customized intention determination model;And during using customizing intention determination model and be customized intention determination model training The intention probability structure of output is intended to order models.
The intention of current customization technical ability is divided into two classes, i.e.,:Customization is intended to and is customized intention;The customization refers to work as The preceding stylish increased intention for being intended to and modifying to obtain to being customized technical ability intention of customization technical ability customization;It is described to be determined The intention that the foundational skills that system refers to use during current customization technical ability customization are not changed in being intended to.It can be seen that it is customized intention It is to be customized the subset being intended in technical ability, is intended to be possible to change in customization because being customized in technical ability, is intended to as customization.
Training customize be intended to determine model when, the saying template being intended to and entity will be customized as positive example training data, The entity is solid data corresponding to semantic groove in saying template, and the solid data is added in corresponding semantic groove, obtained To a positive example training data, collect a large amount of non-customized intentions saying template and entity as counter-example training data, such as can It is intended to true as counter-example training data, training customization using the selected section saying template from the general intention of open platform and entity Cover half type.The input that the customization is intended to determine model is the term vector of positive example training data and counter-example training data, exports and is Each training data belongs to the probability that each customization is intended to.Such as customize technical ability and include 5 intentions, then customization is intended to determine model Export the probability for belonging to this 5 intentions for each training data;The training data is above-mentioned positive example and counter-example data.
Equally, it is customized the training process for being intended to determine model to be intended to determine that the training process of model is similar with customization, i.e., Using be customized intention saying template and entity as positive example training data, collect a large amount of non-saying templates for being customized intention and Entity, such as can selected section saying template and entity conduct from the general intention of open platform as counter-example training data Counter-example training data, training are customized intention and determine model.The intention that is customized determines that the input of model trains number for positive example According to the term vector with counter-example training data;Export and belong to each probability for being customized intention for each training data;Such as it is customized Technical ability include 10 intention, then be customized be intended to determine model output for each training data belong to this 10 intention it is general Rate.
The customization is intended to determine model and be customized to be intended to determine that model can use the classification commonly used in pattern-recognition Model, such as deep neural network model, convolutional neural networks model, supporting vector machine model, specific training method with it is existing Technical ability is identical, will not be described in detail herein.
It is intended to determine model and is customized intention determination model training, it is necessary to mark and customize when structure is intended to order models When, the training data of model output belongs to the intention probability of highest priority in the probability of each intention, other probability it is preferential Level does not limit;Each training data is belonged to input of the probability as order models of each intention, it is general after exporting as sequence Rate, the intention probability of highest priority comes foremost after sequence, and other orders for being intended to probability do not limit.If any 5 customizations It is intended to probability and 10 and is customized intention probability, be then intended to the defeated input of order models as this 15 intention probability, exports and be 15 intention probability after sequence, come the intention probability for highest priority of foremost.
The intention order models can use neural network model, such as convolutional neural networks model, specific training method It is same as the prior art, it will not be described in detail herein.
Step 202, it is intended to determine model and be customized to be intended to determine model using the customization, determines the request data Belong to during customization is intended to the probability being each intended to and be customized the probability being each intended in intention.
Specifically, it is determined that the request data belongs to during customization is intended to the probability being each intended to and is customized each in intention During the probability of intention, directly determine model and be customized to be intended to really using the term vector of the request data as intention is customized The input of cover half type, the request data is respectively obtained according to model output and is belonging respectively to customize the probability being each intended in intention Be customized be intended in the probability that is each intended to.
Step 203, the request data is belonging respectively to customize the probability being each intended in intention and is customized in intention The probability being each intended to inputs the intention order models, the intention after being sorted according to the output of the intention order models Probability.
Step 204, what selected and sorted was most forward is intended to be intended to the intention as the request data corresponding to probability.
Step 205, the request data each saying template corresponding with the intention is matched, obtains the request Saying template corresponding to data.
Specifically, the method that accurate matching or fuzzy matching can be used.
Correspondingly, semantic understanding is carried out to the request data using the saying template, obtains semantic understanding result.Such as use Family request data is " my main points chicken cubes in chilly sauce ", the intention that is intended to order dishes of request data, according to the saying mould for intention of ordering dishes Plate, it may be determined that the semantic groove of active user's request data is " I am main points dish ", and wherein dish is semantic groove, correspondent entity For " chicken cubes in chilly sauce ".
As shown in figure 3, being another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention, wrap Include following steps:
Step 301, previously according to the saying template of the current intention for customizing technical ability, exercise skill is intended to determine model, and The intention probability structure for being intended to determine to export during model training using technical ability is intended to order models.
Specifically, the saying template for being intended to and being customized intention and corresponding entity will be customized and trains number together as positive example According to, a large amount of non-customized intentions and the non-saying template for being customized intention and entity are collected as counter-example training data, such as can be with As counter-example training data, exercise skill is intended to determine for selected section saying template and entity from the general intention of open platform Model.The input that the technical ability is intended to determine model is positive example training data and the term vector of counter-example training data, is exported to be every Individual training data belongs in current customization technical ability the probability being each intended to.
When structure is intended to order models, directly it can be intended to determine the intention that model exports during model training using technical ability Probability, specific training method is similar with said process, that is, when marking technical ability intention determination model training, the output of model is training Data belong to the intention probability of highest priority in the probability of each intention, and the priority of other probability does not limit;Utilize mark Intention probability training afterwards is intended to order models.
Step 302, it is intended to determine model using the technical ability, determines that the request data belongs to every during customization technical ability is intended to The probability of individual intention.
Specifically, it is determined that when the request data belongs to during technical ability is intended to the probability being each intended to, directly by the request The term vector of data is intended to determine the input of model as technical ability, and exporting to obtain the request data according to model belongs to technical ability meaning The probability being each intended in figure.
Step 303, the request data is belonged to during customization technical ability is intended to the probability input intention being each intended to arrange Sequence model, the intention probability after being sorted according to the output of the intention order models.
Step 304, what selected and sorted was most forward is intended to be intended to the intention as the request data corresponding to probability.
Step 305, the request data each saying template corresponding with the intention is matched, obtains the request Saying template corresponding to data.
As shown in figure 4, being another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention, wrap Include following steps:
Step 401, mould is determined previously according to the saying template of the current intention for customizing technical ability, exercise skill saying template Type, and the saying template probabilities for determining to export during model training using technical ability saying template build saying template order models.
Specifically, will current customization technical ability intention saying template and corresponding entity together as positive example training data, The entity is solid data corresponding to semantic groove in saying template, and the solid data is added in corresponding semantic groove, obtained To a positive example training data;Collect the intention of a large amount of non-customized technical ability saying template and entity as counter-example training data, Correspondingly, the solid data is added in corresponding saying template, obtains a counter-example training data, such as can be from opening Selected section saying template and entity determine model as counter-example, exercise skill saying template in the general intention of platform.It is described Technical ability saying template determines that the input of model is positive example training data and the term vector of counter-example training data, exports as each training Data belong to the probability of each saying template in current customization technical ability.
When building saying template order models, model is defeated when directly can determine model training using technical ability saying template The saying template probabilities gone out, specific training method is similar with said process, i.e., when mark technical ability saying template determines model training, The training data of model output belongs to the saying template probabilities of highest priority in the probability of each saying template, other probability Priority does not limit;Saying template order models are trained using the saying template probabilities after mark.
Step 402, model is determined using the technical ability saying template, determines that the request data belongs to the meaning of customization technical ability The probability of each saying template in figure.
Step 403, the request data is belonged to described in the probability input of each saying template in the intention of customization technical ability Saying template order models, the saying template probabilities after being sorted according to the output of the saying template order models.
Step 404, saying template corresponding to the most forward saying template probabilities of selected and sorted is as the request data pair The saying template answered.
As shown in figure 5, being another flow chart that saying template corresponding to request data is determined in the embodiment of the present invention, wrap Include following steps:
Step 501, it is intended to previously according to the saying template that customization is intended to and the saying template for being customized intention, training customization Saying template determines that model determines model with intention saying template is customized;And using customization be intended to saying template determine model and It is customized and is intended to the saying template probabilities structure saying template order models that saying template determines to export during model training.
The intention of current customization technical ability is divided into two classes, i.e.,:Customization is intended to and is customized intention;The customization refers to work as The preceding stylish increased intention for being intended to and modifying to obtain to being customized technical ability intention of customization technical ability customization;It is described to be determined The intention that the foundational skills that system refers to use during current customization technical ability customization are not changed in being intended to.
When training customization intention saying template determines model, the saying template being intended to and entity will be customized and instructed as positive example Practice data, the entity is solid data corresponding to semantic groove in saying template, and the solid data is added into corresponding semanteme In groove, obtain a positive example training data, collect a large amount of non-customized intentions saying template and entity as counter-example training data, Correspondingly, the solid data is added in corresponding saying template, obtains a counter-example training data, such as can be from opening As counter-example training data, it is true that training customization is intended to saying template for selected section saying template and entity in the general intention of platform Cover half type.The customization be intended to saying template determine the input of model for positive example training data and counter-example training data word to Amount, export the probability for belonging to each saying template during customization is intended to for each training data.
Equally, it is customized intention saying template and determines that the training process of model and customization are intended to saying template and determine model Training process is similar, will be customized intention saying template and entity as positive example training data, collect a large amount of non-is customized The saying template and entity of intention are as counter-example training data, for example selector can be defended oneself from the general intention of open platform As counter-example training data, training is customized intention saying template and determines model for method template and entity.The intention that is customized is said Method template determines that the input of model is positive example training data and the term vector of counter-example training data;Export as each training data category In the probability for being customized each saying template in intention.
The customization is intended to saying template and determines that model determines that model can use pattern with intention saying template is customized The disaggregated model commonly used in identification, such as deep neural network model, convolutional neural networks model, supporting vector machine model, tool Body training method is identical with existing technical ability, will not be described in detail herein.
Determine model, it is necessary to mark customization and be intended to saying template when building saying template order models and be customized intention When saying template determines model training, the training data of model output belongs to excellent in the probability of each saying template during technical ability is intended to First level highest saying template probabilities, the priority of other probability do not limit;Each training data is belonged into current customization technical ability Input of the probability of each saying template as order models in intention, the probability after exporting as sequence, priority is most after sequence High saying template probabilities come foremost, and the order of other saying template probabilities does not limit.
The saying template order models can use neural network model, such as convolutional neural networks model, specific training Method is same as the prior art, will not be described in detail herein.
Step 502, it is intended to saying template using the customization and determines that model determines model with intention saying template is customized, Determine that the request data belongs to the probability of each saying template during customization is intended to and is customized each saying template in intention Probability.
Specifically, directly using the term vector of the request data as customization be intended to saying template determine model and by Customization is intended to the input that saying template determines model, and respectively obtaining the request data according to model output is belonging respectively to customization meaning The probability for each saying template being respectively intended in figure and the probability for being customized each saying template being respectively intended in intention.
Step 503, the request data is belonging respectively to the probability of each saying template during customization is intended to and is customized meaning The probability of each saying template inputs the saying template order models in figure, according to the output of the saying template order models Saying template probabilities after being sorted.
Step 504, saying template corresponding to the most forward saying template probabilities of selected and sorted is as the corresponding saying Template.
Semantic understanding method provided in an embodiment of the present invention based on customization technical ability, after user's request data is received, The path of current customization technical ability is obtained, and all semantic resources of the customization technical ability are obtained according to the path, it is determined that described please Ask data to belong to the saying template of current customization technical ability, according to the saying template determine semantic groove in the request data and Correspondent entity, obtain semantic understanding result.Utilize the present invention, it is possible to achieve fast and accurately semantic to user's request data to manage Solution.
Correspondingly, the embodiment of the present invention also provides a kind of semantic understanding system based on customization technical ability, as shown in fig. 6, being A kind of structural representation of the system.
In this embodiment, the system includes:
Receiving module 601, for receiving user's request data;
Source obtaining module 602, for obtaining the path of current customization technical ability, current customization skill is obtained according to the path The semantic resources of energy;
Saying template determining module 603, for determining that the request data belongs to the saying template of current customization technical ability, institute State in method template comprising one or more semantic grooves, each corresponding entity of semantic groove;
Semantic understanding module 604, for determining semantic groove in the request data and correspondingly according to the saying template Entity, obtain semantic understanding result.
Because each technical ability can include multiple intentions, and each it is intended to include a variety of saying templates.Therefore, above-mentioned saying Template determining module 603 can have a variety of realizations when it is determined that the request data belongs to the saying template of current customization technical ability Mode.This is described in detail separately below.
In one embodiment of saying template determining module 603, it can include being intended to determining module and template matches Module.Wherein, the determining module that is intended to is used for the intention for determining that the request data belongs to current customization technical ability;The template Matching module is used to be matched the request data each saying template corresponding with the intention, obtains the request data Corresponding saying template.In this embodiment, it is necessary to first determine that the request data belongs to current by the intention determining module The intention of technical ability is customized, then again as the template matches module by the request data each saying mould corresponding with the intention Plate is matched, and obtains saying template corresponding to the request data.
It should be noted that in actual applications, the determining module that is intended to can determine according to the intention built in advance Model, determine that the request data belongs to the intention of current customization technical ability.Because customization technical ability is that application developer is being customized The technical ability developed in skills base, therefore, the current technical ability that customizes not only include customization technical ability, in addition to are customized technical ability.Accordingly Ground, the current intention for customizing technical ability include:Customization is intended to and is customized intention;The customization refers to currently customize technical ability customization Stylish increased intention and the intention for modifying to obtain to being customized technical ability intention;Described be customized refers to currently determine The intention that the foundational skills used during technical ability customization processed are not changed in being intended to.
Correspondingly, the saying for being intended to determine that model can be the saying template being intended to based on customization and be customized intention The model or the saying template of intention is customized including being based respectively on and is customized saying for intention that template training obtains Two models that method template training obtains.
For example also include in one embodiment of the system:First model construction module (not shown), for advance root According to the saying template (customize the saying template of intention and be customized the saying template of intention) of the intention of current customization technical ability, instruction Practice technical ability to be intended to determine model;And the intention probability structure for being intended to determine to export during model training using technical ability is intended to sequence mould Type.The training method of model is above being described in detail, and will not be repeated here.
Correspondingly, it is described to be intended to determine that a kind of concrete structure of model includes following each unit:
First intention probability determining unit, for being intended to determine model using the technical ability, determine the request data category The probability being each intended in customization technical ability is intended to;
First sequencing unit, institute is inputted for the request data to be belonged to during customization technical ability is intended into the probability being each intended to Intention order models are stated, the intention probability after being sorted according to the output of the intention order models;
First choice unit, most forward for selected and sorted is intended to be intended to as the request data corresponding to probability It is intended to.
For another example, also include in one embodiment of the system:Second model construction module (not shown), for advance According to the saying template that customization is intended to and the saying template for being customized intention, training customization is intended to determine model and is customized intention Determine model;And it is intended to determine that model builds meaning with the intention probability exported when being customized and being intended to determine model training using customization Figure order models.The training method of model is above being described in detail, and will not be repeated here.
Correspondingly, it is described to be intended to determine that a kind of concrete structure of model includes following each unit:
Second intention probability determining unit, for being intended to determine model and be customized to be intended to determine mould using the customization Type, determine that the request data belongs to during customization is intended to the probability being each intended to and is customized the probability being each intended in intention;
Second sequencing unit, for the request data to be belonging respectively to customize the probability being each intended in intention and be determined The probability that system is each intended in being intended to inputs the intention order models, is sorted according to the output of the intention order models Intention probability afterwards;
Second selecting unit, most forward for selected and sorted is intended to be intended to as the request data corresponding to probability It is intended to.
In another embodiment of saying template determining module 603, saying template can be built in advance and determines model, root Directly determine the request data belongs to which saying template of current customization technical ability according to the model.
Similarly, since customization technical ability is the technical ability that application developer is developed on skills base is customized, it is therefore, current fixed Technical ability processed not only includes customization technical ability, in addition to is customized technical ability.Correspondingly, the saying template determines that model can be based on Customize the saying template being intended to and be customized a model that the saying template training of intention obtains or including base respectively The saying template that is intended in customization and it is customized two models that the saying template training of intention obtains.
For example also include in one embodiment of the system:3rd model construction module (not shown), for advance root According to the saying template of the intention of current customization technical ability, exercise skill saying template determines model, and true using technical ability saying template Determine the saying template probabilities structure saying template order models exported during model training.
Correspondingly, a kind of concrete structure of the saying template determining module 603 includes following each unit:
3rd probability determining unit, for determining model using the technical ability saying template, determine the request data category The probability of each saying template in the intention of customization technical ability;
3rd sequencing unit, for the request data to be belonged to the probability of each saying template in the intention for customizing technical ability The saying template order models are inputted, the saying template after being sorted according to the output of the saying template order models is general Rate;
3rd selecting unit, saying template corresponding to the saying template probabilities most forward for selected and sorted are used as described ask Seek saying template corresponding to data.
For another example, also include in one embodiment of the system:4th model construction module (not shown), for advance The saying template that is intended to according to customization and be customized the saying template of intention, training customization be intended to saying template determine model and by Customization is intended to saying template and determines model;And using customization be intended to saying template determine model and be customized be intended to saying template it is true Determine the saying template probabilities structure saying template order models exported during model training.
Correspondingly, a kind of concrete structure of the saying template determining module 603 includes following each unit:
4th probability determining unit, for determining model using the customization intention saying template and being customized intention saying Template determines model, determines that the request data belongs to the probability of each saying template during customization is intended to and is customized every in intention The probability of individual saying template;
4th sequencing unit, for by the request data be belonging respectively to during customization is intended to the probability of each saying template and The probability for being customized each saying template in being intended to inputs the saying template order models, according to saying template sequence mould The output of type sorted after saying template probabilities;
4th selecting unit, saying template is as described right corresponding to the saying template probabilities most forward for selected and sorted The saying template answered.
Semantic understanding system provided in an embodiment of the present invention based on customization technical ability, after user's request data is received, The path of current customization technical ability is obtained, and all semantic resources of the customization technical ability are obtained according to the path, it is determined that described please Ask data to belong to the saying template of current customization technical ability, according to the saying template determine semantic groove in the request data and Correspondent entity, obtain semantic understanding result.Utilize the present invention, it is possible to achieve fast and accurately semantic to user's request data to manage Solution.
Each embodiment in this specification is described by the way of progressive, identical similar portion between each embodiment Divide mutually referring to what each embodiment stressed is the difference with other embodiment.It is moreover, described above System embodiment it is only schematical, wherein the module illustrated as separating component can be or may not be It is physically separate, you can with positioned at a place, or can also to be distributed on multiple NEs.Can be according to reality Need to select some or all of module therein to realize the purpose of this embodiment scheme.Those of ordinary skill in the art are not In the case of paying creative work, you can to understand and implement.
The embodiment of the present invention is described in detail above, embodiment used herein is carried out to the present invention Illustrate, the explanation of above example is only intended to help to understand method and device of the invention;Meanwhile for the one of this area As technical staff, according to the thought of the present invention, there will be changes in specific embodiments and applications, to sum up institute State, this specification content should not be construed as limiting the invention.

Claims (12)

  1. A kind of 1. semantic understanding method based on customization technical ability, it is characterised in that including:
    Receive user's request data;
    The path of current customization technical ability is obtained, the semantic resources of current customization technical ability are obtained according to the path;
    Determine that the request data belongs to the saying template of current customization technical ability, one or more languages are included in the saying template Adopted groove, each corresponding entity of semantic groove;
    The semantic groove and correspondent entity in the request data are determined according to the saying template, obtains semantic understanding result.
  2. 2. according to the method for claim 1, it is characterised in that described to determine that the request data belongs to current customization technical ability Saying template include:
    Determine that the request data belongs to the intention of current customization technical ability;
    The request data each saying template corresponding with the intention is matched, obtains saying corresponding to the request data Method template.
  3. 3. according to the method for claim 2, it is characterised in that methods described also includes:
    Previously according to the saying template of the intention of current customization technical ability, exercise skill is intended to determine model;And it is intended to using technical ability The intention probability structure for determining to export during model training is intended to order models;
    It is described to determine that the request data belongs to the current intention for customizing technical ability and included:
    Using the technical ability be intended to determine model, determine the request data belong to customization technical ability be intended in be each intended to it is general Rate;
    The request data is belonged to during customization technical ability is intended to the probability being each intended to and inputs the intention order models, according to institute State be intended to order models output sorted after intention probability;
    It is intended to the intention as the request data corresponding to the most forward intention probability of selected and sorted.
  4. 4. according to the method for claim 2, it is characterised in that the intention of the current customization technical ability includes:Customization is intended to It is intended to being customized;The customization refers to currently customize the stylish increased intention of technical ability customization and to being customized technical ability meaning Figure modifies obtained intention;It is described to be customized during the foundational skills for referring to be used during current customization technical ability customization are intended to not The intention changed;
    Methods described also includes:
    The saying template that is intended to previously according to customization and be customized the saying template of intention, training customization be intended to determine model and by Customization is intended to determine model;And it is intended to determine that model and the intention exported when being customized and being intended to determine model training are general using customization Rate structure is intended to order models;
    It is described to determine that the request data belongs to the current intention for customizing technical ability and included:
    It is intended to determine model and be customized to be intended to determine model using the customization, determines that the request data belongs to customization and is intended to In the probability that is each intended to and being customized be intended in the probability that is each intended to;
    By the request data be belonging respectively to during customization is intended to the probability that is each intended to and being customized in intention be each intended to it is general Rate inputs the intention order models, the intention probability after being sorted according to the output of the intention order models;
    It is intended to the intention as the request data corresponding to the most forward intention probability of selected and sorted.
  5. 5. according to the method for claim 1, it is characterised in that methods described also includes:
    Previously according to the saying template of the intention of current customization technical ability, exercise skill saying template determines model, and utilizes technical ability Saying template determines the saying template probabilities structure saying template order models exported during model training;
    It is described to determine that the request data belongs to the current saying template for customizing technical ability and included:
    Model is determined using the technical ability saying template, determines that the request data belongs to each saying in the intention for customizing technical ability The probability of template;
    The probability that the request data is belonged to each saying template in the intention of customization technical ability inputs the saying template sequence Model, the saying template probabilities after being sorted according to the output of the saying template order models;
    Saying template is as saying template corresponding to the request data corresponding to the most forward saying template probabilities of selected and sorted.
  6. 6. according to the method for claim 1, it is characterised in that the intention of current customization technical ability includes:Customization be intended to and by Customization is intended to;The customization refer to currently customize technical ability customize it is stylish it is increased be intended to and to be customized technical ability be intended into The intention that row modification obtains;Described be customized during the foundational skills for referring to be used during current customization technical ability customization are intended to is not repaiied The intention changed;
    Methods described also includes:
    Previously according to the saying template that customization is intended to and the saying template for being customized intention, training customization is intended to saying template and determined Model determines model with intention saying template is customized;And it is intended to saying template using customization and determines model and be customized to be intended to say Method template determines the saying template probabilities structure saying template order models exported during model training;
    It is described to determine that the request data belongs to the current saying template for customizing technical ability and included:
    It is intended to saying template using the customization and determines that model determines model with intention saying template is customized, determines the request Data belong to the probability of each saying template and the probability for being customized each saying template in intention during customization is intended to;
    The request data is belonging respectively to the probability of each saying template during customization is intended to and is customized each saying in intention The probability of template inputs the saying template order models, after being sorted according to the output of the saying template order models Saying template probabilities;
    Saying template is as the corresponding saying template corresponding to the most forward saying template probabilities of selected and sorted.
  7. A kind of 7. semantic understanding system based on customization technical ability, it is characterised in that including:
    Receiving module, for receiving user's request data;
    Source obtaining module, for obtaining the path of current customization technical ability, the language of current customization technical ability is obtained according to the path Adopted resource;
    Saying template determining module, for determining that the request data belongs to the saying template of current customization technical ability, the saying One or more semantic grooves, each corresponding entity of semantic groove are included in template;
    Semantic understanding module, for determining semantic groove and correspondent entity in the request data according to the saying template, obtain To semantic understanding result.
  8. 8. system according to claim 7, it is characterised in that the saying template determining module includes:
    It is intended to determining module, for determining that the request data belongs to the intention of current customization technical ability;
    Template matches module, for the request data each saying template corresponding with the intention to be matched, obtain institute State saying template corresponding to request data.
  9. 9. system according to claim 8, it is characterised in that the system also includes:
    First model construction module, for the saying template of the intention previously according to current customization technical ability, exercise skill is intended to true Cover half type;And the intention probability structure for being intended to determine to export during model training using technical ability is intended to order models;
    The intention determining module includes:
    First intention probability determining unit, for being intended to determine model using the technical ability, determine that the request data belongs to fixed The probability that technical ability processed is each intended in being intended to;
    First sequencing unit, the meaning is inputted for the request data to be belonged to during customization technical ability is intended into the probability being each intended to Figure order models, the intention probability after being sorted according to the output of the intention order models;
    First choice unit, most forward for selected and sorted is intended to meaning of the intention as the request data corresponding to probability Figure.
  10. 10. system according to claim 8, it is characterised in that the intention of the current customization technical ability includes:Customization is intended to It is intended to being customized;The customization refers to currently customize the stylish increased intention of technical ability customization and to being customized technical ability meaning Figure modifies obtained intention;It is described to be customized during the foundational skills for referring to be used during current customization technical ability customization are intended to not The intention changed;
    The system also includes:
    Second model construction module, for the saying template being intended to previously according to customization and the saying template for being customized intention, instruction Practice customization to be intended to determine that model determines model with intention is customized;And it is intended to determine model and be customized to be intended to determine using customization The intention probability structure exported during model training is intended to order models;
    The intention determining module includes:
    Second intention probability determining unit, for being intended to determine model and be customized to be intended to determine model using the customization, really The fixed request data belongs to during customization is intended to the probability being each intended to and is customized the probability being each intended in intention;
    Second sequencing unit, for being belonging respectively to customize the probability being each intended in intention by the request data and being customized meaning The probability being each intended in figure inputs the intention order models, after being sorted according to the output of the intention order models It is intended to probability;
    Second selecting unit, most forward for selected and sorted is intended to meaning of the intention as the request data corresponding to probability Figure.
  11. 11. system according to claim 7, it is characterised in that the system also includes:
    3rd model construction module, for the saying template of the intention previously according to current customization technical ability, exercise skill saying mould Plate determines model, and the saying template probabilities for determining to export during model training using technical ability saying template build the sequence of saying template Model;
    The saying template determining module includes:
    3rd probability determining unit, for determining model using the technical ability saying template, determine that the request data belongs to fixed The probability of each saying template in the intention of technical ability processed;
    3rd sequencing unit, the probability for the request data to be belonged to each saying template in the intention for customizing technical ability input The saying template order models, the saying template probabilities after being sorted according to the output of the saying template order models;
    3rd selecting unit, saying template is as the number of request corresponding to the saying template probabilities most forward for selected and sorted According to corresponding saying template.
  12. 12. system according to claim 7, it is characterised in that the intention of current customization technical ability includes:Customization be intended to and by Customization is intended to;The customization refer to currently customize technical ability customize it is stylish it is increased be intended to and to be customized technical ability be intended into The intention that row modification obtains;Described be customized during the foundational skills for referring to be used during current customization technical ability customization are intended to is not repaiied The intention changed;
    The system also includes:
    4th model construction module, for the saying template being intended to previously according to customization and the saying template for being customized intention, instruction Practice customization intention saying template and determine that model determines model with intention saying template is customized;And it is intended to saying template using customization Determine model and be customized to be intended to the saying template probabilities structure saying template sequence that saying template determines to export during model training Model;
    The saying template determining module includes:
    4th probability determining unit, determine model and be customized to be intended to saying template for being intended to saying template using the customization Model is determined, determines that the request data belongs to during customization is intended to the probability of each saying template and is customized in intention and each says The probability of method template;
    4th sequencing unit, for the request data to be belonging respectively into the probability of each saying template during customization is intended to and determined The probability input saying template order models of system each saying template in being intended to, according to the saying template order models Export the saying template probabilities after being sorted;
    4th selecting unit, saying template is as described corresponding corresponding to the saying template probabilities most forward for selected and sorted Saying template.
CN201710774597.9A 2017-08-31 2017-08-31 Semantic understanding method and system based on customization technology Active CN107515857B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710774597.9A CN107515857B (en) 2017-08-31 2017-08-31 Semantic understanding method and system based on customization technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710774597.9A CN107515857B (en) 2017-08-31 2017-08-31 Semantic understanding method and system based on customization technology

Publications (2)

Publication Number Publication Date
CN107515857A true CN107515857A (en) 2017-12-26
CN107515857B CN107515857B (en) 2020-08-18

Family

ID=60724962

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710774597.9A Active CN107515857B (en) 2017-08-31 2017-08-31 Semantic understanding method and system based on customization technology

Country Status (1)

Country Link
CN (1) CN107515857B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108563631A (en) * 2018-03-23 2018-09-21 江苏速度信息科技股份有限公司 A kind of automatic identifying method of natural language address descriptor
CN108984157A (en) * 2018-07-27 2018-12-11 苏州思必驰信息科技有限公司 Technical ability configuration and call method and system for voice dialogue platform
CN110334201A (en) * 2019-07-18 2019-10-15 中国工商银行股份有限公司 A kind of intension recognizing method, apparatus and system
WO2021055247A1 (en) * 2019-09-16 2021-03-25 Oracle International Corporation Stop word data augmentation for natural language processing

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102262622A (en) * 2010-05-31 2011-11-30 国际商业机器公司 Document processing, template generating and conceptbase generating methods and devices
CN102622417A (en) * 2012-02-20 2012-08-01 北京搜狗信息服务有限公司 Method and device for ordering information records
CN102880649A (en) * 2012-08-27 2013-01-16 北京搜狗信息服务有限公司 Individualized information processing method and system
CN104077407A (en) * 2014-07-10 2014-10-01 中国工商银行股份有限公司 System and method for intelligent data searching
CN104199810A (en) * 2014-08-29 2014-12-10 科大讯飞股份有限公司 Intelligent service method and system based on natural language interaction
CN104360994A (en) * 2014-12-04 2015-02-18 科大讯飞股份有限公司 Natural language understanding method and natural language understanding system
US20160004766A1 (en) * 2006-10-10 2016-01-07 Abbyy Infopoisk Llc Search technology using synonims and paraphrasing
CN105487663A (en) * 2015-11-30 2016-04-13 北京光年无限科技有限公司 Intelligent robot oriented intention identification method and system
CN106021339A (en) * 2016-05-09 2016-10-12 中国联合网络通信集团有限公司 A semantic query method and system for a resource tree
CN107015962A (en) * 2017-03-16 2017-08-04 北京光年无限科技有限公司 Towards the implementation method and device of the self-defined intention assessment of intelligent robot

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160004766A1 (en) * 2006-10-10 2016-01-07 Abbyy Infopoisk Llc Search technology using synonims and paraphrasing
CN102262622A (en) * 2010-05-31 2011-11-30 国际商业机器公司 Document processing, template generating and conceptbase generating methods and devices
CN102622417A (en) * 2012-02-20 2012-08-01 北京搜狗信息服务有限公司 Method and device for ordering information records
CN102880649A (en) * 2012-08-27 2013-01-16 北京搜狗信息服务有限公司 Individualized information processing method and system
CN104077407A (en) * 2014-07-10 2014-10-01 中国工商银行股份有限公司 System and method for intelligent data searching
CN104199810A (en) * 2014-08-29 2014-12-10 科大讯飞股份有限公司 Intelligent service method and system based on natural language interaction
CN104360994A (en) * 2014-12-04 2015-02-18 科大讯飞股份有限公司 Natural language understanding method and natural language understanding system
CN105487663A (en) * 2015-11-30 2016-04-13 北京光年无限科技有限公司 Intelligent robot oriented intention identification method and system
CN106021339A (en) * 2016-05-09 2016-10-12 中国联合网络通信集团有限公司 A semantic query method and system for a resource tree
CN107015962A (en) * 2017-03-16 2017-08-04 北京光年无限科技有限公司 Towards the implementation method and device of the self-defined intention assessment of intelligent robot

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108563631A (en) * 2018-03-23 2018-09-21 江苏速度信息科技股份有限公司 A kind of automatic identifying method of natural language address descriptor
CN108984157A (en) * 2018-07-27 2018-12-11 苏州思必驰信息科技有限公司 Technical ability configuration and call method and system for voice dialogue platform
CN110334201A (en) * 2019-07-18 2019-10-15 中国工商银行股份有限公司 A kind of intension recognizing method, apparatus and system
WO2021055247A1 (en) * 2019-09-16 2021-03-25 Oracle International Corporation Stop word data augmentation for natural language processing

Also Published As

Publication number Publication date
CN107515857B (en) 2020-08-18

Similar Documents

Publication Publication Date Title
CN107515857A (en) Semantic understanding method and system based on customization technical ability
CN106407178A (en) Session abstract generation method and device
CN105912692B (en) A kind of method and apparatus of Intelligent voice dialog
CN106406806A (en) A control method and device for intelligent apparatuses
CN108769772B (en) Direct broadcasting room display methods, device, equipment and storage medium
CN109074802A (en) The modulation of packetizing audio signal
CN105045919B (en) A kind of information output method and device
CN107644002A (en) A kind of searching articles method, apparatus and system
CN105491126A (en) Service providing method and service providing device based on artificial intelligence
CN108959531A (en) Information search method, device, equipment and storage medium
CN108521366A (en) Expression method for pushing and electronic equipment
CN106356056B (en) Audio recognition method and device
CN110992124A (en) House resource recommendation method and system
CN110162675A (en) Generation method, device, computer-readable medium and the electronic equipment of answer statement
CN109582822A (en) A kind of music recommended method and device based on user speech
CN109408800A (en) Talk with robot system and associative skills configuration method
CN108564494A (en) Teaching resource promotes sharing means
CN108140055A (en) Trigger application message
CN107560617A (en) A kind of air navigation aid based on advertisement machine, storage medium and advertisement machine
CN107450725A (en) Man-machine interaction application platform, method and storage medium
CN105389333B (en) A kind of searching system construction method and server architecture
CN108334353B (en) Skill development system and method
CN110176227A (en) A kind of method and relevant apparatus of speech recognition
CN109739969A (en) Answer generation method and intelligent conversational system
CN108776685A (en) Automatic question-answering method, server and storage medium

Legal Events

Date Code Title Description
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