WO2022001682A1 - 一种车载***的控件对象查询方法和装置 - Google Patents

一种车载***的控件对象查询方法和装置 Download PDF

Info

Publication number
WO2022001682A1
WO2022001682A1 PCT/CN2021/100617 CN2021100617W WO2022001682A1 WO 2022001682 A1 WO2022001682 A1 WO 2022001682A1 CN 2021100617 W CN2021100617 W CN 2021100617W WO 2022001682 A1 WO2022001682 A1 WO 2022001682A1
Authority
WO
WIPO (PCT)
Prior art keywords
word
target
words
control object
candidate entity
Prior art date
Application number
PCT/CN2021/100617
Other languages
English (en)
French (fr)
Inventor
张又亮
申众
张崇宇
Original Assignee
广州橙行智动汽车科技有限公司
广州小鹏汽车科技有限公司
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 广州橙行智动汽车科技有限公司, 广州小鹏汽车科技有限公司 filed Critical 广州橙行智动汽车科技有限公司
Publication of WO2022001682A1 publication Critical patent/WO2022001682A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
    • G06F3/0481—Interaction techniques based on graphical user interfaces [GUI] based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295—Named entity recognition

Definitions

  • the present application relates to the technical field of vehicles, and in particular, to a method for querying a control object of an in-vehicle system and a device for querying a control object of an in-vehicle system.
  • the embodiments of the present application are proposed to provide an in-vehicle system control object query method and a corresponding in-vehicle system control object inquiry device that overcomes or at least partially solves the above problems.
  • the embodiment of the present application discloses a control object query method for an in-vehicle system, including:
  • the knowledge graph records a plurality of entity words and an edge connecting two entity words based on the modification relationship of the entity words, and the entity words include an action word, a control object word and attribute information word;
  • the connected graph includes at least two of the target entity words, and an edge connected between the two target entity words that matches the modification relationship of the query keyword;
  • selecting a target entity word matching the query keyword from the entity words recorded in the knowledge graph including:
  • the feature parameters of a single candidate entity word and a preset machine learning model are used to sort multiple candidate entity words for a single query keyword, and the target entity word is selected according to the sorting result.
  • calculating the feature parameters of a single candidate entity word according to candidate entity words corresponding to multiple query keywords including:
  • connection degree of the single candidate entity word calculates the average connection degree of the single candidate entity word
  • selecting a target path from the path of the connectivity graph including:
  • selecting the target path that contains the most target entity words and the shortest path from the path of the connected graph including:
  • the target path with the most target entity words and the shortest path is selected.
  • the entity words recorded in the knowledge graph also include page attribute words; the determining, according to the target path, the target page where the control object corresponding to the control object word is located, including:
  • the method is applied to a server; the method further includes:
  • the embodiment of the present application also discloses a device for querying a control object of an in-vehicle system, including:
  • a user query statement processing module is used to obtain a user query statement, and extract query keywords and the modification relationship of the query keywords from the user query statement, where the query keywords include application words, control object words and attribute information words. ;
  • the knowledge graph acquisition module is used to obtain the knowledge graph constructed for the vehicle-mounted system, the knowledge graph records a plurality of entity words and the edge connecting the two entity words based on the modification relationship of the entity words, and the entity words include the actuation verb , control object words and attribute information words;
  • a target entity word selection module configured to select a target entity word matching the query keyword from the entity words recorded in the knowledge map
  • the connected graph search module is used to search the connected graph from the knowledge graph;
  • the connected graph includes at least two of the target entity words, and a modification connected between the two target entity words and the query keyword The edges of the relationship match;
  • a target path selection module for selecting a target path from the path of the connected graph
  • the control information generation module is used to determine the target page where the control object corresponding to the control object word is located according to the target path and generate a jump instruction, and to determine the display position of the control object corresponding to the control object word on the target page and generating position information; so that the in-vehicle system jumps to the target page according to the jump instruction, and slides to the display position on the target page for display according to the position information.
  • the target entity word selection module includes:
  • a candidate entity word selection submodule is used to select a corresponding candidate entity word for the query keyword from the entity words recorded in the knowledge map;
  • a feature parameter calculation submodule configured to calculate the feature parameters of a single candidate entity word according to candidate entity words corresponding to multiple query keywords
  • the target entity word selection sub-module is used for selecting the target entity word from the candidate entity words for a single query keyword by using the characteristic parameter of the single candidate entity word.
  • the characteristic parameter calculation submodule includes:
  • a side number determining unit used for respectively determining the number of sides between the single candidate entity word corresponding to each query keyword and the candidate entity words corresponding to other query keywords;
  • an average number of sides calculation unit used for calculating the average number of sides of the single candidate entity word by using the number of sides between the single candidate entity word and the candidate entity words corresponding to other query keywords
  • a connectivity determination unit configured to determine the connectivity of the single candidate word by using the number of edges between the single candidate entity word and the candidate entity words corresponding to other query keywords
  • connection degree average determining unit configured to use the connection degree of the single candidate entity word to calculate the connection degree average value of the single candidate entity word
  • the sorting unit is used to determine the similarity between the query keyword and the corresponding candidate entity words, and sort each candidate entity word according to the similarity, so as to obtain a single candidate entity word in multiple entities corresponding to the same query keyword.
  • the feature parameter determination unit is used to determine the average number of edges of the single candidate entity word, the average connection degree of the single candidate entity word, and the single candidate entity word in multiple candidate entity words corresponding to the same query keyword. At least one of the ranking positions in , as the feature parameter of the candidate entity word.
  • the target path selection module includes:
  • the target path selection sub-module is used to select the target path containing the most target entity words and the shortest path from the paths of the connected graph.
  • the target path selection submodule includes:
  • an available control information obtaining unit used for obtaining the available control information of the vehicle-mounted system
  • a path deletion unit configured to delete a path that does not match the available control information from the path of the connectivity graph
  • the target path selection unit is used for selecting the target path containing the most target entity words and the shortest path in the remaining paths.
  • the entity words recorded in the knowledge graph also include page attribute words;
  • the control information generation module includes:
  • a page attribute word search submodule used to search for a page attribute word that has a modification relationship with the control object word from the target path;
  • the target page determination submodule is used for determining the target page corresponding to the page attribute word.
  • the apparatus is applied to a server; the apparatus further includes:
  • a control information sending module is used to send the jump instruction and the position information to the vehicle, so that the vehicle's on-board system jumps to the target page according to the jump instruction, and according to the position information
  • the target page is slid to the display position of the target page where the widget object corresponding to the widget object word is displayed.
  • An embodiment of the present application further discloses an electronic device, including: a processor, a memory, and a computer program stored on the memory and capable of running on the processor, the computer program being implemented when executed by the processor The steps of the control object query method of the in-vehicle system as described above.
  • the embodiment of the present application further discloses a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned method for querying a control object of an in-vehicle system are implemented .
  • the knowledge graph constructed for the vehicle-mounted system can be obtained; from the entity words recorded in the knowledge graph, select the target entity word that matches the query keyword extracted from the user query sentence; each of the target entity words, and a connection graph connecting the edges matching the modification relationship of the query keyword between the two target entity words; selecting a target path from the path of the connection graph; determining the corresponding control object words according to the target path
  • the information is displayed by sliding to the display position on the target page.
  • the user can quickly lock the page and the controls only by describing his own needs through language, thereby reducing the user search cost and improving the user experience. And by matching the entity words instantiated in the knowledge graph, the accuracy of identifying user query sentences can be improved, and the query efficiency can be improved.
  • FIG. 1 is a flow chart of steps of an embodiment of a method for querying a control object of an in-vehicle system of the present application
  • Fig. 2 is a schematic diagram of the modification relationship of query keywords and query keywords extracted from user query sentences in the present application
  • 3 is a schematic diagram representing the connection relationship between candidate entity words in the present application.
  • Fig. 4 is the schematic diagram of a kind of target path in the present application.
  • FIG. 5 is a structural block diagram of an embodiment of an apparatus for querying a control object of an in-vehicle system according to the present application.
  • the present application proposes a method for direct access to scene elements based on the knowledge map of the vehicle system, which can accurately understand user query statements, and can help users jump to pages that need to be set. Users only need to describe their needs through language.
  • Knowledge graph is essentially a semantic network, which is a graph-based data structure composed of nodes (Point) and edges (Edge).
  • each node represents an "entity” existing in the real world
  • each edge is a "relationship” between entities.
  • FIG. 1 a flow chart of steps of an embodiment of a method for querying a control object of an in-vehicle system of the present application is shown, and the method may specifically include the following steps:
  • Step 101 Obtain a user query statement, and extract query keywords and modification relationships of the query keywords from the user query statement, where the query keywords include an application word, a control object word, and an attribute information word.
  • ASR Automatic Speech Recognition
  • the scene elements of the in-vehicle system directly belong to task-type dialogues, mainly imperative sentences, most of which are verb-object phrases, which need to clarify the entities of the controls and the actions to be applied.
  • Dependency syntax considers that the verb in the "predicate" is the center of a sentence, and other components are directly or indirectly related to the verb, which can determine the predicate and the corresponding entity.
  • the present application is directed to the direct access of spatial objects in in-vehicle systems.
  • the query keyword may include a control object word and an action verb, the action verb is a predicate, and the control object word is the entity imposed by the action verb.
  • FIG. 2 it is a schematic diagram of extracting query keywords from a user query sentence and the modification relationship of the query keywords in the present application.
  • the arrow points indicate the modifier relationship, for example, the arrow points from “open” to “page”, indicating that "open” has a modifier relationship to "page”, such as verb-object relationship.
  • the arrow points from “Settings” to “Font”, indicating that “Settings” has a modifier relationship to "Font”, such as verb-object relationship.
  • the arrow points from "font” to "page”, indicating that "font” points to "page” with a modifier relationship.
  • Attribute information words are words for attribute descriptions of entities, including attribute values. For example, in “media volume is set to twenty”, “twenty” is the attribute value of "media volume”.
  • Step 102 Obtain a knowledge graph constructed for the vehicle-mounted system, where the knowledge graph records a plurality of entity words and an edge connecting two entity words based on the modification relationship of the entity words, and the entity words include an actuation verb and a control object word. and attribute information words.
  • the knowledge graph constructed for the in-vehicle system may be the original resource data for the in-vehicle system in advance.
  • the original resource data is divided into structured data, semi-structured data and unstructured data according to the structure.
  • the original resource data may include general vehicle configuration data, vehicle remote maintenance record information, sales store information and other business data. It can also include personalized data for in-vehicle applications, such as data uploaded to the server by the Graphical User Interface (GUI) of in-vehicle applications, design documents for in-vehicle applications, user manuals, real user query statements, and other related data.
  • GUI Graphical User Interface
  • the entity nodes of the knowledge graph store entity words, and the edge connecting two entity nodes represents the modification relationship of the entity words of the two entity nodes.
  • Entity words are keywords extracted from the original resource data based on the control direct scene of the in-vehicle system, which can include action words, control object words and attribute information words.
  • Step 103 From the entity words recorded in the knowledge graph, select a target entity word that matches the query keyword.
  • the target entity words that match the query keywords can be the same words, or can be generalized to have the same meaning.
  • the step of selecting a target entity word matching the query keyword from the entity words recorded in the knowledge graph may include the following sub-steps:
  • Sub-step S11 from the entity words recorded in the knowledge graph, select a corresponding candidate entity word for the query keyword.
  • the candidate entity words can be the same words as the query keywords, or can be generalized to have the same meaning.
  • the generation method of candidate entity words is mainly based on the method of "string comparison" based on the surface form of entity referents and the entity names in the knowledge base.
  • a named dictionary template may be employed that contains representations of the names of various entities, such as: variants, abbreviations, obfuscated names, spelling variants, and nicknames. With the expansion of the magnitude of the entity, it can also be searched by technology based on search engines.
  • Sub-step S12 according to the candidate entity words corresponding to the plurality of query keywords, calculate the characteristic parameters of a single candidate entity word.
  • the sub-step S12 may further include:
  • Sub-step S121 Determine the number of edges between each candidate entity word corresponding to each query keyword and each candidate entity word corresponding to another query keyword.
  • the candidate entity word lists 1 to 4 are respectively lists of candidate entity words corresponding to four query keywords, and each candidate entity word list may include four candidate entity words.
  • the candidate entity word 1 of the candidate entity word list 1 is connected to the candidate entity word 2 of the candidate entity word list 2, and the number of edges is 1. Connect with candidate entity word 2 of candidate entity word list 3, and the number of edges is 1. It is connected with candidate entity word 2 of candidate entity word list 4, and the number of edges is 1.
  • Sub-step S122 using the number of edges between the single candidate entity word and the candidate entity words corresponding to other query keywords, to calculate the average number of edges of the single candidate entity word.
  • the number of edges between the candidate entity word 1 in the candidate entity word list 1 and the candidate entity word in the candidate entity word list 2 is 1, and the number of edges between the candidate entity word in the candidate entity word list 3 and the candidate entity word in the candidate entity word list 3 is 1.
  • the number of edges with the candidate entity word in the candidate entity word list 4 is 1, so the average value of the edge of the candidate entity word 1 in the candidate entity word list 1 is 1.
  • connection degree of the single candidate word is determined by using the number of edges between the single candidate entity word and the candidate entity words corresponding to other query keywords.
  • the degree of connectivity is another measure of the number of edges of candidate entity words. If the number of edges between a single candidate entity word and each candidate entity word corresponding to another query keyword is greater than or equal to the preset number of edges threshold, then the connection is made. degree is 1; if not, the degree of connectivity is 0.
  • the preset edge number threshold may be 2.
  • the connection degree is 0.
  • the number of edges between the candidate entity word 1 in the candidate entity word list 1 and the candidate entity word in the candidate entity word list 2 is 1, so the connection degree is 0.
  • the number of edges between the candidate entity word 1 in the candidate entity word list 1 and the candidate entity word in the candidate entity word list 3 is 1, so the connection degree is 0.
  • connection degree is 0.
  • Sub-step S124 using the connection degree of the single candidate entity word to calculate the average connection degree of the single candidate entity word.
  • Sub-step S125 Determine the similarity between the query keyword and the corresponding candidate entity words, and sort each candidate entity word according to the similarity to obtain a single candidate entity word in multiple candidates corresponding to the same query keyword. Sort position in entity words.
  • Sub-step S126 sorting the average number of edges of the single candidate entity word, the average connection degree of the single candidate entity word, and the single candidate entity word among multiple candidate entity words corresponding to the same query keyword. At least one of the positions is used as a feature parameter of the candidate entity word.
  • Sub-step S13 using the feature parameters of a single candidate entity word and a preset machine learning model to sort multiple candidate entity words for a single query keyword, and select a target entity word according to the sorting result.
  • each candidate entity word can be input into a pre-trained machine learning model, for example, the xgboost (eXtreme Gradient Boosting, extreme gradient boosting algorithm) model, and the score of the candidate entity word can be obtained; the candidate entity word with the highest score is selected as target entity word.
  • a pre-trained machine learning model for example, the xgboost (eXtreme Gradient Boosting, extreme gradient boosting algorithm) model
  • Step 104 from the knowledge graph, find a connectivity graph; the connectivity graph includes at least two of the target entity words, and an edge connected between the two target entity words that matches the modification relationship of the query keyword .
  • a connected graph is a graph connected by entity nodes and edges. There may be multiple edges between two target entity words in the knowledge graph, and different edges correspond to different modification relationships. In this application, it is necessary to search for the modification relationship that contains at least two target entity words and is connected to the edge between the two target entity words from the knowledge graph, and needs to match the modification relationship of the query keyword.
  • Step 105 Select a target path from the paths in the connectivity graph.
  • the knowledge graph there may be a plurality of connected graphs including at least two target entity words and having edges connected between the two target entity words that match the modification relationship of the query keyword.
  • the path of a connected graph consists of entity nodes and edges in the connected graph.
  • the step of selecting a target path from the paths of the connectivity graph may include: selecting a target path that contains the most target entity words and has the shortest path from the paths of the connectivity graph.
  • the number of target entity words contained in each path can be counted, and the path containing the most target entity words can be screened to ensure that as many target entity words as possible are covered in the path.
  • the redundant information of non-user query statements in the path is avoided, and the shortest path is screened out.
  • the control objects supported by the in-vehicle system are different.
  • the step of selecting the target path containing the most target entity words and the shortest path from the path of the connectivity graph may include: acquiring available control information of the vehicle-mounted system; from the path of the connectivity graph Delete the path that does not match the available control information; in the remaining paths, select the target path that contains the most target entity words and has the shortest path.
  • the control object that the user wants to query may not be supported by the in-vehicle system.
  • the available control information records the control object supported by the in-vehicle system. If the control object recorded in the path of the Unicom map does not belong to the control object supported by the in-vehicle system, the Path removed.
  • Step 106 determine the target page where the control object corresponding to the control object word is located and generate a jump instruction, and determine the display position of the control object corresponding to the control object word on the target page and generate position information; So that the in-vehicle system jumps to the target page according to the jump instruction, and slides the target page to the display position for display according to the location information.
  • the method of the present application can be applied to a server, and can also be directly applied to an in-vehicle system.
  • the server sends the jump instruction and the location information to the vehicle, so that the on-board system of the vehicle jumps to the target page according to the jump instruction, and according to the location information
  • the target page is slid to the display position of the target page where the widget object corresponding to the widget object word is displayed.
  • a page attribute word is a word that describes whether an entity is a page, for example setting the "Page" entity to indicate that another entity is a page.
  • FIG. 4 a schematic diagram of a target path in the present application is shown.
  • the user query statement is "open the page where the font is set", and the corresponding entity nodes of the target path include: “OpenAction”, “Page”, “Display Settings”, “Font Size”, “Tablayout”, “Setaction”, among which "OpenAction” " is the entity definition for the open action, "Page” is the entity definition for the page class, “Tablayout” is the entity definition for the option list, and “Setaction” is the entity definition for the set action.
  • the relationship includes "action” to indicate action, “instanceOf” to indicate that the specified entity belongs to a certain type of instance relationship, and “has” to indicate the ownership relationship between entity node A and entity node B.
  • “Page” points to "Display Settings”, and the relationship is “instanceOf”, indicating that "Display Settings” belongs to the page class.
  • “Tablayout” points to "Font Size”, and the relationship is “instanceOf”, indicating that "Font Size” belongs to the list of options.
  • "Display Settings” points to "Font Size”, and the relationship "has” means that the list of options for "Font Size” is in the Display Settings page.
  • the application can obtain the knowledge graph constructed for the vehicle system; from the entity words recorded in the knowledge graph, select the target entity words that match the query keywords extracted from the user query sentence; The target entity word, and the connection graph of the edge that is connected between the two target entity words and the modification relationship of the query keyword; from the path of the connection graph, select the target path; determine the corresponding control object word according to the target path.
  • This application only requires users to describe their needs through language, and the page and controls can be quickly locked, reducing user search costs and improving user experience. And by matching the entity words instantiated in the knowledge graph, the accuracy of identifying user query sentences can be improved, and the query efficiency can be improved.
  • FIG. 5 a structural block diagram of an embodiment of a device for querying a control object of an in-vehicle system of the present application is shown, which may specifically include the following modules:
  • the user query statement processing module 501 is configured to obtain a user query statement, and extract query keywords and the modification relationship of the query keywords from the user query statement, where the query keywords include application verbs, control object words and attribute information word;
  • the knowledge graph obtaining module 502 is used to obtain the knowledge graph constructed for the vehicle-mounted system.
  • the knowledge graph records a plurality of entity words and the edge connecting the two entity words based on the modification relationship of the entity words. verbs, control object words and attribute information words;
  • a target entity word selection module 503, configured to select a target entity word matching the query keyword from the entity words recorded in the knowledge graph;
  • the connected graph search module 504 is configured to search for a connected graph from the knowledge graph; the connected graph includes at least two of the target entity words, and a connection between the two target entity words and the query keyword. Decorate the edges that match the relationship;
  • the target path selection module 505 is used to select the target path from the path of the connectivity graph
  • the control information generation module 506 is used to determine the target page where the control object corresponding to the control object word is located according to the target path, generate a jump instruction, and determine the display position of the control object corresponding to the control object word on the target page and generate position information; so that the in-vehicle system jumps to the target page according to the jump instruction, and slides the target page to the display position for display according to the position information.
  • the target entity word selection module 503 may include:
  • a candidate entity word selection submodule is used to select a corresponding candidate entity word for the query keyword from the entity words recorded in the knowledge map;
  • a feature parameter calculation submodule configured to calculate the feature parameters of a single candidate entity word according to candidate entity words corresponding to multiple query keywords
  • the target entity word selection sub-module is configured to use the feature parameters of a single candidate entity word and a preset machine learning model to sort multiple candidate entity words for a single query keyword, and select the target entity word according to the sorting result.
  • the characteristic parameter calculation sub-module may include:
  • a side number determining unit used for respectively determining the number of sides between the single candidate entity word corresponding to each query keyword and the candidate entity words corresponding to other query keywords;
  • an average number of sides calculation unit used for calculating the average number of sides of the single candidate entity word by using the number of sides between the single candidate entity word and the candidate entity words corresponding to other query keywords
  • a connectivity determination unit configured to determine the connectivity of the single candidate word by using the number of edges between the single candidate entity word and the candidate entity words corresponding to other query keywords
  • connection degree average determining unit configured to use the connection degree of the single candidate entity word to calculate the connection degree average value of the single candidate entity word
  • the sorting unit is used to determine the similarity between the query keyword and the corresponding candidate entity words, and sort each candidate entity word according to the similarity, so as to obtain a single candidate entity word in multiple entities corresponding to the same query keyword.
  • the feature parameter determination unit is used to determine the average number of edges of the single candidate entity word, the average connection degree of the single candidate entity word, and the single candidate entity word in multiple candidate entity words corresponding to the same query keyword. At least one of the ranking positions in , as the feature parameter of the candidate entity word.
  • the target path selection module 505 may include:
  • the target path selection sub-module is used to select the target path containing the most target entity words and the shortest path from the paths of the connected graph.
  • the target path selection submodule may include:
  • an available control information obtaining unit used for obtaining the available control information of the vehicle-mounted system
  • a path deletion unit configured to delete a path that does not match the available control information from the path of the connectivity graph
  • the target path selection unit is used for selecting the target path containing the most target entity words and the shortest path in the remaining paths.
  • the entity words recorded in the knowledge graph also include page attribute words; the control information generation module 506 may include:
  • a page attribute word search submodule used to search for a page attribute word that has a modification relationship with the control object word from the target path;
  • the target page determination submodule is used for determining the target page corresponding to the page attribute word.
  • the device is applied to a server; the device further includes:
  • a control information sending module is used to send the jump instruction and the position information to the vehicle, so that the vehicle's on-board system jumps to the target page according to the jump instruction, and according to the position information
  • the target page is slid to the display position of the target page where the widget object corresponding to the widget object word is displayed.
  • the embodiment of the present application also provides an electronic device, including:
  • Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, each process of the above embodiment of the method for querying a control object of an in-vehicle system is implemented, and can achieve The same technical effect, in order to avoid repetition, will not be repeated here.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (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)
  • Human Computer Interaction (AREA)
  • Navigation (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

本申请公开了一种车载***的控件对象查询方法和装置，所述方法包括：获取用户查询语句，并从用户查询语句提取查询关键词以及查询关键词的修饰关系；获取针对车载***构建的知识图谱；从知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；从知识图谱中，查找联通图；从联通图的路径中，选取目标路径；根据目标路径确定控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定控件对象词对应的控件对象在目标页面的显示位置并生成位置信息；以使车载***根据跳转指令跳转至目标页面，以及根据位置信息在目标页面滑动至显示位置进行。本申请实施例可以降低用户查找成本，提升用户体验。

Description

一种车载***的控件对象查询方法和装置
本申请要求在2020年06月30日提交中国专利局、申请号202010615618.4、发明名称为“一种车载***的控件对象查询方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及车辆技术领域，特别是涉及一种车载***的控件对象查询方法和一种车载***的控件对象查询装置。
背景技术
随着智能汽车的快速发展，为满足用户的需求，在车载***里面会搭载各种各样的应用程序APP，如导航、空调、音乐、***设置等，每个应用程序包含大量页面，每个页面包含大量可操作性的控件。针对复杂的图形用户界面GUI***，用户通过点击或者手势常用的操作方式，无法快速找到自己需要操作的控件。
车载***可操作性的控件繁多，如果用常规的分类方法进行场景直达，需要的确定的意图比较多，没办法一一枚举；而且训练模型需要的大量泛化数据和标注，成本较高。
发明内容
鉴于上述问题，提出了本申请实施例以便提供一种克服上述问题或者至少部分地解决上述问题的一种车载***的控件对象查询方法和相应的一种车载***的控件对象查询装置。
本申请实施例公开了一种车载***的控件对象查询方法，包括：
获取用户查询语句，并从所述用户查询语句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词；
获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动 词、控件对象词和属性信息词；
从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；
从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边；
从所述联通图的路径中，选取目标路径；
根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
可选地，所述从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词，包括：
从所述知识图谱记录的实体词中，针对所述查询关键词选取对应的候选实体词；
根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数；
采用单个所述候选实体词的特征参数和预设机器学习模型，对针对单个查询关键词的多个候选实体词进行排序，并根据排序结果选取目标实体词。
可选地，所述根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数，包括：
分别确定各个查询关键词对应的单个候选实体词与其他查询关键词对应的候选实体词之间的边数；
采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，计算所述单个候选实体词的边数平均值；
采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，确定所述单个候选词的连接度；
采用所述单个候选实体词的连接度，计算所述单个候选实体词的连接度平均值；
确定所述查询关键词与对应的各个候选实体词之间的相似度，并根据相似度对各个候选实体词进行排序，得到单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置；
将所述单个候选实体词的边数平均值、所述单个候选实体词的连接度平均值、所述单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置中的至少一种，作为所述候选实体词的特征参数。
可选地，所述从所述联通图的路径中，选取目标路径，包括：
从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径。
可选地，所述从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径，包括：
获取所述车载***的可用控件信息；
从所述联通图的路径中删除与所述可用控件信息不匹配的路径；
在剩余的路径中，选取包含的目标实体词最多且路径最短的目标路径。
可选地，所述知识图谱记录的实体词还包括页面属性词；所述根据目标路径，确定所述控件对象词对应的控件对象所在的目标页面，包括：
从所述目标路径中，查找与所述控件对象词具有修饰关系的页面属性词；
确定所述页面属性词对应的目标页面。
可选地，所述的方法应用于服务器；所述方法还包括：
向车辆发送所述跳转指令和所述位置信息，以使所述车辆的车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述控件对象词对应的控件对象在所述目标页面的显示位置进行显示。
本申请实施例还公开了一种车载***的控件对象查询装置，包括：
用户查询语句处理模块，用于获取用户查询语句，并从所述用户查询语 句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词；
知识图谱获取模块，用于获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动词、控件对象词和属性信息词；
目标实体词选取模块，用于从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；
联通图查找模块，用于从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边；
目标路径选取模块，用于从所述联通图的路径中，选取目标路径；
控制信息生成模块，用于根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
可选地，所述目标实体词选取模块包括：
候选实体词选取子模块，用于从所述知识图谱记录的实体词中，针对所述查询关键词选取对应的候选实体词；
特征参数计算子模块，用于根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数；
目标实体词选取子模块，用于采用单个所述候选实体词的特征参数，在针对单个查询关键词的候选实体词中，选取目标实体词。
可选地，所述特征参数计算子模块包括：
边数确定单元，用于分别确定各个查询关键词对应的单个候选实体词与其他查询关键词对应的候选实体词之间的边数；
边数平均值计算单元，用于采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，计算所述单个候选实体词的边数平均值；
连接度确定单元，用于采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，确定所述单个候选词的连接度；
连接度平均值确定单元，用于采用所述单个候选实体词的连接度，计算所述单个候选实体词的连接度平均值；
排序单元，用于确定所述查询关键词与对应的各个候选实体词之间的相似度，并根据相似度对各个候选实体词进行排序，得到单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置；
特征参数确定单元，用于将所述单个候选实体词的边数平均值、所述单个候选实体词的连接度平均值、所述单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置中的至少一种，作为所述候选实体词的特征参数。
可选地，所述目标路径选取模块包括：
目标路径选取子模块，用于从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径。
可选地，所述目标路径选取子模块包括：
可用控件信息获取单元，用于获取所述车载***的可用控件信息；
路径删除单元，用于从所述联通图的路径中删除与所述可用控件信息不匹配的路径；
目标路径选取单元，用于在剩余的路径中，选取包含的目标实体词最多且路径最短的目标路径。
可选地，所述知识图谱记录的实体词还包括页面属性词；所述控制信息生成模块包括：
页面属性词查找子模块，用于从所述目标路径中，查找与所述控件对象词具有修饰关系的页面属性词；
目标页面确定子模块，用于确定所述页面属性词对应的目标页面。
可选地，所述的装置应用于服务器；所述装置还包括：
控制信息发送模块，用于向车辆发送所述跳转指令和所述位置信息，以使所述车辆的车载***根据所述跳转指令跳转至所述目标页面，以及根据所 述位置信息在所述目标页面滑动至所述控件对象词对应的控件对象在所述目标页面的显示位置进行显示。
本申请实施例还公开了一种电子设备，包括：处理器、存储器及存储在所述存储器上并能够在所述处理器上运行的计算机程序，所述计算机程序被所述处理器执行时实现如上所述的车载***的控件对象查询方法的步骤。
本申请实施例还公开了一种计算机可读存储介质，所述计算机可读存储介质上存储计算机程序，所述计算机程序被处理器执行时实现如上所述的车载***的控件对象查询方法的步骤。
本申请实施例包括以下优点：
本申请实施例，可以获取针对车载***构建的知识图谱；从知识图谱记录的实体词中，选取与从用户查询语句提取的查询关键词匹配的目标实体词；从知识图谱中，查找包含至少两个所述目标实体词，以及连接在两个目标实体词之间与查询关键词的修饰关系匹配的边的联通图；从联通图的路径中，选取目标路径；根据目标路径确定控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定控件对象词对应的控件对象在目标页面的显示位置并生成位置信息；以使车载***根据跳转指令跳转至目标页面，以及根据位置信息在目标页面滑动至显示位置进行显示。本申请实施例只需要用户通过语言描述自己的需求，即可快速锁定页面以及控件，降低用户查找成本，提升用户体验。并且通过对知识图谱中实例化的实体词进行匹配，可以提高对用户查询语句识别的准确度，提高查询效率。
附图说明
本申请上述的和/或附加的方面和优点从下面结合附图对实施例的描述中将变得明显和容易理解，其中：
图1是本申请的一种车载***的控件对象查询方法实施例的步骤流程图；
图2是本申请中从用户查询语句提取查询关键词以及查询关键词的修饰关系的示意图；
图3是本申请中表示候选实体词之间的连接关系的示意图；
图4是本申请中一种目标路径的示意图；
图5是本申请的一种车载***的控件对象查询装置实施例的结构框图。
具体实施方式
本申请提出一种基于车载***知识图谱的场景元素直达的方法，能精确理解用户查询语句，可以帮助用户跳转需要设置的页面，用户只需要通过语言描述自己的需求即可。
知识图谱本质上是语义网络，是一种基于图的数据结构，由节点(Point)和边(Edge)组成。在知识图谱里，每个节点表示现实世界中存在的“实体”，每条边为实体与实体之间的“关系”。
参照图1，示出了本申请的一种车载***的控件对象查询方法实施例的步骤流程图，所述方法具体可以包括如下步骤：
步骤101，获取用户查询语句，并从所述用户查询语句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词。
可以利用自动语音识别ASR(Automatic Speech Recognition)对用户语音进行识别，得到用户查询语句。
车载***的场景元素直达属于任务型的对话，以祈使句为主，大部分为动宾短语，需要明确控件的实体，以及施加的动作。依存句法认为“谓语”中的动词是一个句子的中心，其他成分与动词直接或间接地产生联系，可以确定谓语和对应的实体。
本申请针对的是车载***中空间对象的直达。查询关键词可以包括控件对象词和施动词，施动词即谓语，控件对象词是施动词施加的实体。
利用句法依存分析，可以分析查询语句中词语之间的关系。例如，利用基于转移的List-based Arc-eager算法，确定句子成分，获取句子成分间的有向无环图。参照图2所示为本申请中从用户查询语句提取查询关键词以及查询关键词的修饰关系的示意图。
针对用户查询语句“打开设置字体页面”，获取施动词“打开”、“设置”和控件对象词“字体”、“页面”。
获取查询关键词之间的修饰成分；在图中，箭头指向表示修饰关系，例 如，箭头从“打开”指向“页面”，表示“打开”针对“页面”具有修饰关系，如动宾关系。箭头从“设置”指向“字体”，表示“设置”针对“字体”具有修饰关系，如动宾关系。箭头从“字体”指向“页面”，表示“字体”指向“页面”具有修饰关系。
属性信息词是针对实体的属性描述的词，包括属性值。例如，在“媒体音量设为二十”中，“二十”为“媒体音量”的属性值。
步骤102，获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动词、控件对象词和属性信息词。
针对车载***构建的知识图谱，可以是预先采用针对车载***的原始资源数据。原始资源数据按结构划分，可以分为的结构化数据、半结构化数据和非结构化数据。原始资源数据可以包括通用的车辆配置数据、车辆远程维修记录信息、销售门店信息等业务数据。还可以包括针对车载应用的个性化数据，例如，车载应用的图形用户界面GUI(Graphical User Interface)上传到服务器的数据，针对车载应用的设计文档，使用手册，真实的用户查询语句，以及其他相关数据。
知识图谱的实体节点中存储有实体词，连接两个实体节点的边表示两个实体节点的实体词的修饰关系。
实体词是基于车载***的控件直达场景从原始资源数据中抽取得到的关键词，可以包括施动词、控件对象词和属性信息词。
步骤103，从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词。
与查询关键词匹配的目标实体词可以是相同的词语，也可以是被归纳为具有相同含义的词语。
在本申请中，所述从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词的步骤可以包括如下子步骤：
子步骤S11，从所述知识图谱记录的实体词中，针对所述查询关键词选取对应的候选实体词。
候选实体词可以是与查询关键词相同的词语，也可以是被归纳为具有相同含义的词语。
候选实体词的生成方法主要是针对实体指称项表层形式和知识库中的实体名称，进行基于“字串比较”的方式而生成。在一种示例中，可以采用命名字典模板，命名字典模板包含各种实体的名称的表达方式，如：变体、缩写、混淆名称、拼写变体和昵称等。随着实体量级的扩展，还可以采用基于搜索引擎的技术查找。
子步骤S12，根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数。
所述子步骤S12可以进一步包括：
子步骤S121，确定各个查询关键词对应的单个候选实体词，分别与另一个查询关键词对应的各个候选实体词之间的边数。
参照图3所示为本申请中表示候选实体词之间的连接关系的示意图。其中，候选实体词列表1-4，分别是对应4个查询关键词的候选实体词的列表，每个候选实体词列表可以包括4个候选实体词。
候选实体词列表1的候选实体词1，与候选实体词列表2的候选实体词2连接，边数为1。与候选实体词列表3的候选实体词2连接，边数为1。与候选实体词列表4的候选实体词2连接，边数为1。
子步骤S122，采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，计算所述单个候选实体词的边数平均值。
如图3所示，候选实体词列表1的候选实体词1，与候选实体词列表2中的候选实体词的边数为1，与候选实体词列表3中的候选实体词的边数为1，与候选实体词列表4中的候选实体词的边数为1，因此候选实体词列表1的候选实体词1的边平均值为1。
子步骤S123，采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，确定所述单个候选词的连接度。
连接度是对候选实体词的边数的另一种衡量标准，若单个候选实体词与另一个查询关键词对应的各个候选实体词之间的边数大于或等于预设边数 阈值，则连接度为1；若否，则连接度为0。在一种示例中，预设边数阈值可以为2。
如图3所示，候选实体词列表1的候选实体词1，与候选实体词列表2中的候选实体词的边数为1，因此其连接度为0。
候选实体词列表1的候选实体词1，与候选实体词列表2中的候选实体词的边数为1，因此其连接度为0。
候选实体词列表1的候选实体词1，与候选实体词列表3中的候选实体词的边数为1，因此其连接度为0。
候选实体词列表1的候选实体词1，与候选实体词列表4中的候选实体词的边数为1，因此其连接度为0。
子步骤S124，采用所述单个候选实体词的连接度，计算所述单个候选实体词的连接度平均值。
子步骤S125，确定所述查询关键词与对应的各个候选实体词之间的相似度，并根据相似度对各个候选实体词进行排序，得到单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置。
子步骤S126，将所述单个候选实体词的边数平均值、所述单个候选实体词的连接度平均值、所述单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置中的至少一种，作为所述候选实体词的特征参数。
子步骤S13，采用单个所述候选实体词的特征参数和预设机器学习模型，对针对单个查询关键词的多个候选实体词进行排序，并根据排序结果选取目标实体词。
可以将各个候选实体词的特征参数输入预训练的机器学习模型，例如，xgboost(eXtreme Gradient Boosting，极值梯度提升算法)模型，得到该候选实体词的得分；将得分最高的候选实体词选取为目标实体词。
步骤104，从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边。
联通图是由实体节点和边连接的图。知识图谱中两个目标实体词之间的 边可能多个，不同的边对应不同的修饰关系。本申请中，需要从知识图谱中，查找包含至少两个目标实体词，且连接在两个目标实体词之间边的修饰关系，需要与查询关键词的修饰关系匹配。
步骤105，从所述联通图的路径中，选取目标路径。
知识图谱中，可能具有多个包含至少两个目标实体词，且具有连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边的联通图。
联通图的路径由联通图中实体节点和边构成。
在本申请中，所述从所述联通图的路径中，选取目标路径的步骤可以包括：从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径。
具体的，可以统计每条路径中包含目标实体词的数量，筛选出包含目标实体词最多的路径，确保路径中尽可能多的覆盖目标实体词。根据最短路径的原则，避免路径中出现非用户查询语句的冗余信息，筛选出最短的路径。
对于不同车型的车辆，车载***支持的控件对象各有差异。进一步的，所述从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径的步骤可以包括：获取所述车载***的可用控件信息；从所述联通图的路径中删除与所述可用控件信息不匹配的路径；在剩余的路径中，选取包含的目标实体词最多且路径最短的目标路径。
用户想要查询的控件对象可能车载***并不支持，可用控件信息记录了车载***所支持的控件对象，如果联通图的路径中记录的控件对象不属于车载***所支持的控件对象，则将该路径删除。
步骤106，根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
本申请的方法可以应用于服务器，也可以直接应用于车载***。当应用于服务器时，服务器向车辆发送所述跳转指令和所述位置信息，以使所述车 辆的车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述控件对象词对应的控件对象在所述目标页面的显示位置进行显示。
在本申请中，所述知识图谱记录的实体词还包括页面属性词；所述根据目标路径，确定所述控件对象词对应的控件对象所在的目标页面的步骤可以包括：从所述目标路径中，查找与所述控件对象词具有修饰关系的页面属性词；确定所述页面属性词对应的目标页面。
页面属性词是描述实体是否为页面的词，例如设置“Page”实体表示另一个实体为页面。参照图4所示为本申请中一种目标路径的示意图。
用户查询语句为“打开设置字体的页面”，对应的目标路径的实体节点包括：“OpenAction”、“Page”、“显示设置”、“字体大小”、“Tablayout”、“Setaction”，其中“OpenAction”是对打开动作的实体定义，“Page”是对页面类的实体定义、“Tablayout”是对选项列表的实体定义，“Setaction”是对设置动作的实体定义。
关系包括“action”表示动作，“instanceOf”表示指定实体属于某类型的实例关系，“has”表示实体节点A与实体节点B之间的拥有关系。
其中，“Page”指向“显示设置”，且关系为“instanceOf”，表示“显示设置”属于页面类。“Tablayout”指向“字体大小”，且关系为“instanceOf”，表示“字体大小”属于选项列表。“显示设置”指向“字体大小”，且关系为“has”表示“字体大小”的选项列表在“显示设置”的页面中。
本申请可以获取针对车载***构建的知识图谱；从知识图谱记录的实体词中，选取与从用户查询语句提取的查询关键词匹配的目标实体词；从知识图谱中，查找包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边的联通图；从联通图的路径中，选取目标路径；根据目标路径确定控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定控件对象词对应的控件对象在目标页面的显示位置并生成位置信息；以使车载***根据跳转指令跳转至目标页面，以及根据位置信息在目标页面滑动至显示位置进行显示。本申请只需要用户通过语言 描述自己的需求，即可快速锁定页面以及控件，降低用户查找成本，提升用户体验。并且通过对知识图谱中实例化的实体词进行匹配，可以提高对用户查询语句识别的准确度，提高查询效率。
需要说明的是，对于方法实施例，为了简单描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本申请实施例并不受所描述的动作顺序的限制，因为依据本申请实施例，某些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作并不一定是本申请实施例所必须的。
参照图5，示出了本申请的一种车载***的控件对象查询装置实施例的结构框图，具体可以包括如下模块：
用户查询语句处理模块501，用于获取用户查询语句，并从所述用户查询语句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词；
知识图谱获取模块502，用于获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动词、控件对象词和属性信息词；
目标实体词选取模块503，用于从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；
联通图查找模块504，用于从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边；
目标路径选取模块505，用于从所述联通图的路径中，选取目标路径；
控制信息生成模块506，用于根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
在本申请中，所述目标实体词选取模块503可以包括：
候选实体词选取子模块，用于从所述知识图谱记录的实体词中，针对所述查询关键词选取对应的候选实体词；
特征参数计算子模块，用于根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数；
目标实体词选取子模块，用于采用单个所述候选实体词的特征参数和预设机器学习模型，对针对单个查询关键词的多个候选实体词进行排序，并根据排序结果选取目标实体词。
在本申请中，所述特征参数计算子模块可以包括：
边数确定单元，用于分别确定各个查询关键词对应的单个候选实体词与其他查询关键词对应的候选实体词之间的边数；
边数平均值计算单元，用于采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，计算所述单个候选实体词的边数平均值；
连接度确定单元，用于采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，确定所述单个候选词的连接度；
连接度平均值确定单元，用于采用所述单个候选实体词的连接度，计算所述单个候选实体词的连接度平均值；
排序单元，用于确定所述查询关键词与对应的各个候选实体词之间的相似度，并根据相似度对各个候选实体词进行排序，得到单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置；
特征参数确定单元，用于将所述单个候选实体词的边数平均值、所述单个候选实体词的连接度平均值、所述单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置中的至少一种，作为所述候选实体词的特征参数。
在本申请中，所述目标路径选取模块505可以包括：
目标路径选取子模块，用于从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径。
在本申请中，所述目标路径选取子模块可以包括：
可用控件信息获取单元，用于获取所述车载***的可用控件信息；
路径删除单元，用于从所述联通图的路径中删除与所述可用控件信息不匹配的路径；
目标路径选取单元，用于在剩余的路径中，选取包含的目标实体词最多且路径最短的目标路径。
在本申请中，所述知识图谱记录的实体词还包括页面属性词；所述控制信息生成模块506可以包括：
页面属性词查找子模块，用于从所述目标路径中，查找与所述控件对象词具有修饰关系的页面属性词；
目标页面确定子模块，用于确定所述页面属性词对应的目标页面。
在本申请中，所述的装置应用于服务器；所述装置还包括：
控制信息发送模块，用于向车辆发送所述跳转指令和所述位置信息，以使所述车辆的车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述控件对象词对应的控件对象在所述目标页面的显示位置进行显示。
对于装置实施例而言，由于其与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
本申请实施例还提供了一种电子设备，包括：
包括处理器、存储器及存储在所述存储器上并能够在所述处理器上运行的计算机程序，该计算机程序被处理器执行时实现上述车载***的控件对象查询方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供了一种计算机可读存储介质，计算机可读存储介质上存储计算机程序，计算机程序被处理器执行时实现上述车载***的控件对象查询方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，程序可存储于一非易失性计算机可读存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，存储介质可为磁碟、光盘、只读存储记忆体(Read-Only  Memory，ROM)等。
以上实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对本申请专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (10)

  1. 一种车载***的控件对象查询方法，其特征在于，包括：
    获取用户查询语句，并从所述用户查询语句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词；
    获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动词、控件对象词和属性信息词；
    从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；
    从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边；
    从所述联通图的路径中，选取目标路径；
    根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
  2. 根据权利要求1所述的方法，其特征在于，所述从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词，包括：
    从所述知识图谱记录的实体词中，针对所述查询关键词选取对应的候选实体词；
    根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数；
    采用单个所述候选实体词的特征参数和预设机器学习模型，对针对单个查询关键词的多个候选实体词进行排序，并根据排序结果选取目标实体词。
  3. 根据权利要求2所述的方法，其特征在于，所述根据多个查询关键词对应的候选实体词，计算单个所述候选实体词的特征参数，包括：
    分别确定各个查询关键词对应的单个候选实体词与其他查询关键词对应的候选实体词之间的边数；
    采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，计算所述单个候选实体词的边数平均值；
    采用单个候选实体词与其他查询关键词对应的候选实体词之间的边数，确定所述单个候选词的连接度；
    采用所述单个候选实体词的连接度，计算所述单个候选实体词的连接度平均值；
    确定所述查询关键词与对应的各个候选实体词之间的相似度，并根据相似度对各个候选实体词进行排序，得到单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置；
    将所述单个候选实体词的边数平均值、所述单个候选实体词的连接度平均值、所述单个候选实体词在对应同一查询关键词的多个候选实体词中的排序位置中的至少一种，作为所述候选实体词的特征参数。
  4. 根据权利要求1所述的方法，其特征在于，所述从所述联通图的路径中，选取目标路径，包括：
    从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径。
  5. 根据权利要求4所述的方法，其特征在于，所述从所述联通图的路径中，选取包含的目标实体词最多且路径最短的目标路径，包括：
    获取所述车载***的可用控件信息；
    从所述联通图的路径中删除与所述可用控件信息不匹配的路径；
    在剩余的路径中，选取包含的目标实体词最多且路径最短的目标路径。
  6. 根据权利要求1所述的方法，其特征在于，所述知识图谱记录的实体词还包括页面属性词；所述根据目标路径，确定所述控件对象词对应的控件对象所在的目标页面，包括：
    从所述目标路径中，查找与所述控件对象词具有修饰关系的页面属性词；
    确定所述页面属性词对应的目标页面。
  7. 根据权利要求1所述的方法，其特征在于，所述的方法应用于服务器；所述方法还包括：
    向车辆发送所述跳转指令和所述位置信息，以使所述车辆的车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述控件对象词对应的控件对象在所述目标页面的显示位置进行显示。
  8. 一种车载***的控件对象查询装置，其特征在于，包括：
    用户查询语句处理模块，用于获取用户查询语句，并从所述用户查询语句提取查询关键词以及所述查询关键词的修饰关系，所述查询关键词包括施动词、控件对象词和属性信息词；
    知识图谱获取模块，用于获取针对车载***构建的知识图谱，所述知识图谱记录有多个实体词以及基于所述实体词的修饰关系连接两个实体词的边，所述实体词包括施动词、控件对象词和属性信息词；
    目标实体词选取模块，用于从所述知识图谱记录的实体词中，选取与所述查询关键词匹配的目标实体词；
    联通图查找模块，用于从所述知识图谱中，查找联通图；所述联通图包含至少两个所述目标实体词，以及连接在两个目标实体词之间与所述查询关键词的修饰关系匹配的边；
    目标路径选取模块，用于从所述联通图的路径中，选取目标路径；
    控制信息生成模块，用于根据目标路径确定所述控件对象词对应的控件对象所在的目标页面并生成跳转指令，以及确定所述控件对象词对应的控件对象在所述目标页面的显示位置并生成位置信息；以使所述车载***根据所述跳转指令跳转至所述目标页面，以及根据所述位置信息在所述目标页面滑动至所述显示位置进行显示。
  9. 一种电子设备，其特征在于，包括：处理器、存储器及存储在所述存储器上并能够在所述处理器上运行的计算机程序，所述计算机程序被所述处理器执行时实现如权利要求1-7中任一项所述的车载***的控件对象查询 方法的步骤。
  10. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质上存储计算机程序，所述计算机程序被处理器执行时实现如权利要求1至7中任一项所述的车载***的控件对象查询方法的步骤。
PCT/CN2021/100617 2020-06-30 2021-06-17 一种车载***的控件对象查询方法和装置 WO2022001682A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010615618.4A CN111736738A (zh) 2020-06-30 2020-06-30 一种车载***的控件对象查询方法和装置
CN202010615618.4 2020-06-30

Publications (1)

Publication Number Publication Date
WO2022001682A1 true WO2022001682A1 (zh) 2022-01-06

Family

ID=72653839

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/100617 WO2022001682A1 (zh) 2020-06-30 2021-06-17 一种车载***的控件对象查询方法和装置

Country Status (2)

Country Link
CN (1) CN111736738A (zh)
WO (1) WO2022001682A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111736738A (zh) * 2020-06-30 2020-10-02 广州小鹏车联网科技有限公司 一种车载***的控件对象查询方法和装置
CN112634888A (zh) * 2020-12-11 2021-04-09 广州橙行智动汽车科技有限公司 语音交互方法、服务器、语音交互***和可读存储介质
CN112882679B (zh) * 2020-12-21 2022-07-01 广州橙行智动汽车科技有限公司 一种语音交互的方法和装置
CN112579600B (zh) * 2020-12-21 2024-07-19 广州橙行智动汽车科技有限公司 一种基于车载问答的数据处理方法和装置
CN112685535A (zh) * 2020-12-25 2021-04-20 广州橙行智动汽车科技有限公司 语音交互方法、服务器、语音交互***和存储介质
CN112765288A (zh) * 2021-02-05 2021-05-07 新华智云科技有限公司 知识图谱的构建方法及***、信息查询方法及***
CN114121013A (zh) * 2021-12-07 2022-03-01 杭州逗酷软件科技有限公司 语音控制方法、装置、电子设备及存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140298216A1 (en) * 2013-03-28 2014-10-02 The Boeing Company Visualization of an Object Using a Visual Query System
CN107944025A (zh) * 2017-12-12 2018-04-20 北京百度网讯科技有限公司 信息推送方法和装置
CN108776684A (zh) * 2018-05-25 2018-11-09 华东师范大学 知识图谱中边权重的优化方法、装置、介质、设备及***
CN109933653A (zh) * 2019-01-24 2019-06-25 平安科技（深圳）有限公司 问答***的问答查询方法、***及计算机设备
CN110659013A (zh) * 2018-06-28 2020-01-07 比亚迪股份有限公司 一种消息处理方法与装置、存储介质
CN111736738A (zh) * 2020-06-30 2020-10-02 广州小鹏车联网科技有限公司 一种车载***的控件对象查询方法和装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9911417B2 (en) * 2016-04-01 2018-03-06 Tai-An Lu Internet of things system with voice-controlled functions and method for processing information of the same
CN109213844B (zh) * 2018-08-13 2023-03-21 腾讯科技（深圳）有限公司 一种文本处理方法、装置以及相关设备
CN110321408B (zh) * 2019-05-30 2023-07-14 广东省智湾汇科技有限公司 基于知识图谱的搜索方法、装置、计算机设备和存储介质
CN111008532B (zh) * 2019-12-12 2023-09-12 广州小鹏汽车科技有限公司 语音交互方法、车辆和计算机可读存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140298216A1 (en) * 2013-03-28 2014-10-02 The Boeing Company Visualization of an Object Using a Visual Query System
CN107944025A (zh) * 2017-12-12 2018-04-20 北京百度网讯科技有限公司 信息推送方法和装置
CN108776684A (zh) * 2018-05-25 2018-11-09 华东师范大学 知识图谱中边权重的优化方法、装置、介质、设备及***
CN110659013A (zh) * 2018-06-28 2020-01-07 比亚迪股份有限公司 一种消息处理方法与装置、存储介质
CN109933653A (zh) * 2019-01-24 2019-06-25 平安科技（深圳）有限公司 问答***的问答查询方法、***及计算机设备
CN111736738A (zh) * 2020-06-30 2020-10-02 广州小鹏车联网科技有限公司 一种车载***的控件对象查询方法和装置

Also Published As

Publication number Publication date
CN111736738A (zh) 2020-10-02

Similar Documents

Publication Publication Date Title
WO2022001682A1 (zh) 一种车载***的控件对象查询方法和装置
US9613166B2 (en) Search suggestions of related entities based on co-occurrence and/or fuzzy-score matching
JP6657124B2 (ja) 会話理解システムのためのセッションコンテキストモデリング
US9646606B2 (en) Speech recognition using domain knowledge
US9239875B2 (en) Method for disambiguated features in unstructured text
KR100682897B1 (ko) 사전 업데이트 방법 및 그 장치
US20150154197A1 (en) Method for obtaining search suggestions from fuzzy score matching and population frequencies
US20150074112A1 (en) Multimedia Question Answering System and Method
TWI645303B (zh) 字串驗證方法、字串擴充方法與驗證模型訓練方法
US20180075013A1 (en) Method and system for automating training of named entity recognition in natural language processing
JP2018077858A (ja) 会話式情報検索システムおよび方法
US20150100308A1 (en) Automated Formation of Specialized Dictionaries
JP6205477B2 (ja) ビジネスエンティティ用の地理的ロケールのデータの非決定性曖昧さ排除と質的エンティティマッチングのためのシステム
JP2005182280A (ja) 情報検索システム、検索結果加工システム及び情報検索方法並びにプログラム
CN112189229A (zh) 针对计算机化个人助手的技能发现
JP2015511746A5 (zh)
JP2005222532A (ja) 大規模な電子的なドキュメントのコレクション（集まり）を探索するためのドキュメント関連性の決定に対する機械学習アプローチ
US20220414137A1 (en) Automatic labeling of text data
US11429792B2 (en) Creating and interacting with data records having semantic vectors and natural language expressions produced by a machine-trained model
US20090112845A1 (en) System and method for language sensitive contextual searching
Gygli et al. Efficient object annotation via speaking and pointing
AU2022304683A1 (en) Automatic labeling of text data
US9507834B2 (en) Search suggestions using fuzzy-score matching and entity co-occurrence
JP7098502B2 (ja) 報告書作成装置、方法、およびプログラム
JP3787310B2 (ja) キーワード決定方法、装置、プログラム、および記録媒体

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21833274

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 17.05.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21833274

Country of ref document: EP

Kind code of ref document: A1