CN113919360A

CN113919360A - Semantic understanding method, voice interaction method, device, equipment and storage medium

Info

Publication number: CN113919360A
Application number: CN202010666265.0A
Authority: CN
Inventors: 邓憧
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2020-07-09
Filing date: 2020-07-09
Publication date: 2022-01-11

Abstract

The embodiment of the application provides a semantic understanding method, a voice interaction device, electronic equipment and a computer storage medium, and relates to the technical field of computers. Wherein the method comprises the following steps: performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding. Through the embodiment of the application, semantic understanding of the target text in the service field can be effectively enhanced.

Description

Semantic understanding method, voice interaction method, device, equipment and storage medium

Technical Field

The embodiment of the invention relates to the technical field of computers, in particular to a semantic understanding method, a voice interaction method, a device, electronic equipment and a computer storage medium.

Background

Natural Language Processing (NLP) is an important direction in the fields of computer science and artificial intelligence, and enables effective communication between people and computers using natural Language. Natural language processing is a science integrating linguistics, computer science and mathematics. Therefore, the field relates to natural language, namely the language used by people daily, so that the field is closely related to linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic question and answer, knowledge mapping, and the like. Semantic understanding is an important ring of natural language processing technology, and how to accurately understand the requirements of users is the development trend of artificial intelligence technology.

In the conversation process in the vertical field (such as the video field, the music field and the navigation field), sentences of users have strong diversity, the users often mix with some common sense information in the sentences, for example, the common sense information mixed in the sentence "play a movie of grandma of zhang is" grandma of zhang ", and the wrong interpretation of the common sense information in the conversation system often makes the natural language understanding of the whole sentence wrong, thereby causing conversation failure and causing poor experience for the users. Therefore, how to effectively enhance the semantic understanding of the target text in the business field becomes a technical problem to be solved urgently at present.

Disclosure of Invention

In view of the above, embodiments of the present invention provide a semantic understanding scheme to at least partially solve the above technical problems.

According to a first aspect of embodiments of the present invention, a semantic understanding method is provided. The method comprises the following steps: performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

According to a second aspect of the embodiments of the present invention, a voice interaction method is provided. The method comprises the following steps: performing text recognition on received user voice data to obtain a target text to be semantically understood; performing semantic analysis on the target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; and generating response voice data aiming at the user voice data based on the semantic understanding result.

According to a third aspect of embodiments of the present invention, there is provided a semantic understanding method. The method comprises the following steps: receiving a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood; performing semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; and generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

According to a fourth aspect of the embodiments of the present invention, there is provided a semantic understanding apparatus. The device comprises: the first semantic analysis module is used for performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file so as to obtain an entity relation to be replaced in the target text to be semantically understood; the first determination module is used for determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph; a first replacement module, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; and the first semantic understanding module is used for performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

According to a fifth aspect of the embodiments of the present invention, a voice interaction apparatus is provided. The device comprises: the text recognition module is used for performing text recognition on the received user voice data to obtain a target text to be semantically understood; the second semantic analysis module is used for carrying out semantic analysis on the target text to be semantically understood based on the configured semantic analysis file so as to obtain the entity relationship to be replaced in the target text to be semantically understood; a fourth determining module, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced; the second replacement module is used for replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity so as to obtain a replaced target text; the third semantic understanding module is used for performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; and the first generation module is used for generating response voice data aiming at the user voice data based on the semantic understanding result.

According to a sixth aspect of the embodiments of the present invention, there is provided a semantic understanding apparatus. The device comprises: the first receiving module is used for receiving a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood; the third semantic analysis module is used for carrying out semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file so as to obtain an entity relation to be replaced in the target text to be semantically understood; a fifth determining module, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced; a third replacing module, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; the fourth semantic understanding module is used for performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; and the second generation module is used for generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

According to a seventh aspect of the embodiments of the present invention, there is provided an electronic apparatus including: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus; the memory is used for storing at least one executable instruction, which causes the processor to execute the operation corresponding to the semantic understanding method according to the first aspect, or execute the operation corresponding to the voice interaction method according to the second aspect, or execute the operation corresponding to the semantic understanding method according to the third aspect.

According to an eighth aspect of embodiments of the present invention, there is provided a computer storage medium having stored thereon a computer program which, when executed by a processor, implements the semantic understanding method according to the first aspect, or implements the voice interaction method according to the second aspect, or implements the semantic understanding method according to the third aspect.

According to the semantic understanding scheme provided by the embodiment of the invention, the semantic analysis is carried out on the target text to be semantically understood based on the configured semantic analysis file so as to obtain the entity relationship to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the embodiments of the present invention, and it is also possible for a person skilled in the art to obtain other drawings based on the drawings.

Fig. 1A is a schematic flowchart of a robot interacting with a user according to an embodiment of the present disclosure;

FIG. 1B is a flowchart illustrating steps of a semantic understanding method according to an embodiment of the present disclosure;

fig. 1C is a scene schematic diagram of a semantic understanding method according to an embodiment of the present disclosure;

FIG. 2A is a flowchart illustrating steps of a semantic understanding method according to a second embodiment of the present application;

fig. 2B is a scene schematic diagram of a semantic understanding method according to a second embodiment of the present application;

FIG. 3 is a flowchart illustrating steps of a voice interaction method according to a third embodiment of the present application;

FIG. 4 is a flowchart illustrating steps of a semantic understanding method according to a fourth embodiment of the present disclosure;

fig. 5 is a schematic structural diagram of a semantic understanding apparatus according to a fifth embodiment of the present application;

fig. 6 is a schematic structural diagram of a semantic understanding apparatus according to a sixth embodiment of the present application;

fig. 7 is a schematic structural diagram of a voice interaction apparatus according to a seventh embodiment of the present application;

fig. 8 is a schematic structural diagram of a semantic understanding apparatus according to an eighth embodiment of the present application;

fig. 9 is a schematic structural diagram of an electronic device in a ninth embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments obtained by a person skilled in the art based on the embodiments of the present invention shall fall within the scope of the protection of the embodiments of the present invention.

The following further describes specific implementation of the embodiments of the present invention with reference to the drawings.

Artificial intelligence is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

With the continuous evolution of artificial intelligence technology, its application scene begins to emerge constantly in recent years, and more users begin to come into direct contact with intelligent service products, for example common intelligent house, intelligent wearing equipment, virtual assistant, intelligent audio amplifier, intelligent marketing, unmanned, autopilot, unmanned aerial vehicle, robot, intelligent medical treatment, intelligent customer service etc.. Specifically, taking a robot as an example, fig. 1A is a schematic flow chart of interaction between the robot and a user, and as shown in fig. 1A, the user "how is the weather today? "wake up the robot, then the robot converts the user speech into user text by using the speech recognition module, and then performs natural language processing on the user text by using the semantic understanding module to convert the user text into a semantic structure that can be understood by a corresponding service (such as a weather service), for example, a domain: weather; intention is: inquiring the temperature; parameters are as follows: { date: today }, and generates a response result, e.g., weather: sunny; temperature: 20 to 30 ℃; parameters are as follows: { date: today, the response result is converted into response voice by a voice synthesis module, and the response voice is fed back to the user. Among them, the natural language processing technology is an important direction in the fields of computer science and artificial intelligence.

The scheme provided by the embodiment of the application relates to an artificial intelligence natural language processing technology, in particular to a semantic understanding method, a voice interaction device, electronic equipment and a computer storage medium.

Referring to fig. 1B, a flowchart of the steps of the semantic understanding method according to the first embodiment of the present application is shown.

Specifically, the semantic understanding method provided by this embodiment includes the following steps:

in step S101, performing semantic parsing on a target text to be semantically understood based on a configured semantic parsing file to obtain an entity relationship to be replaced in the target text to be semantically understood.

In this embodiment, the semantic parsing file may be understood as a file configured to record semantic extraction rules of a text. Different semantic parsing files can be configured according to different services. Specifically, the semantic analysis file may be a semantic analysis file configured for a video service, a semantic analysis file configured for a music service, or a semantic analysis file configured for a navigation service. The target text to be semantically understood can be a question to be semantically understood, a statement sentence to be semantically understood, or a dialogue between a user and a machine. The semantic analysis can be understood as semantic extraction rules based on the text, and semantic extraction is carried out on the target text to be semantically understood. An entity may refer to a name of an object stored in a knowledge graph, such as a person's name, place name, concept, drug name, company name, and the like. Such as Zhang three, Shanghai, vitamin C, etc. The entity relationship can be understood as another entity that has an associative relationship with the entity in the target text, for example, when the entity is "Zhang three", the entity relationship can be "Ladies of Zhang three", "children of Zhang three", "mom of Zhang three", or "dad of Zhang three". For another example, when the entity is a "third-generation tenth-generation peach blossom", the entity relationship may be a "theme song of the third-generation tenth-generation peach blossom". For another example, when the entity is "Changan twelve hours", the entity relationship may be "the leading actor on the Changan twelve hours". It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, the semantic parsing file comprises an entity tag profile and a semantic template associated with the entity tag profile. Wherein, the entity label configuration file can be understood as a file configured to record entity label matching rules related to the semantic analysis rules. When performing semantic analysis on a target text to be semantically understood based on a configured semantic analysis file, performing named entity identification on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood; replacing an entity corresponding to the named entity tag matched with the entity tag configuration file in the target text to be semantically understood with a preset character; and performing semantic analysis on the target text which is replaced by the preset characters by using a semantic template associated with the entity tag configuration file to obtain the entity relationship to be replaced in the target text to be understood semantically. Wherein the preset character can be a placeholder. Therefore, through the configured entity label configuration file and the configured semantic template associated with the entity label configuration file, the semantic analysis of the target text can be quickly expanded to a new service scene. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In one specific example, the entity may be "Zhang San", "Sansheng SanshiShi Shi Tao", etc., and the entity label may be "singer", "actor", "politician", etc. The entity tag configuration file comprises a file name and at least one entity tag matching rule. The entity tag matching rule may be understood as a rule in an entity tag profile for matching with tags of entities in the target text. For example, the beginning entity tag of the target text part is "singer" and the ending entity tag of the target text part is "actor", the entity tag "actor" exists in the target text, and so on. The semantic template associated with the entity tag profile may be understood as a semantic template having a template name identical to the file name of the entity tag profile. The semantic template comprises regular expressions carrying the preset characters, e.g., "@ (his/her/his/her)? (daughter's son's) "," @ (grandpa's | her | father)? (father son daughter) "where" @ "is a placeholder and the rest are regular expressions. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, when performing named entity recognition on the target text to be semantically understood, performing named entity recognition on the target text to be semantically understood through a named entity recognition model to obtain a named entity tag of the target text to be semantically understood. Therefore, the named entity recognition model is used for recognizing the named entity of the target text to be semantically understood, and the named entity label of the target text to be semantically understood can be accurately recognized. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In one specific example, the named entity recognition model can be a conditional random field model based on a two-way long-short term memory network. The model has high accuracy on the task of sequence annotation. The bidirectional long-short term memory network consists of two ordinary recurrent neural networks, namely a forward recurrent neural network which utilizes the past information, a reverse recurrent neural network which utilizes the future information, so that at the time t, the information at the time t-1 can be used, and the information at the time t can be usedThe information up to time t +1 can be used. Generally, since the bidirectional long and short term memory network can simultaneously use the information of the past time and the future time, the final prediction is more accurate than that of the unidirectional long and short term memory network. The conditional random field model is characterized in that sufficient features with different dimensions are extracted according to massive feature engineering, and then sequence labeling is carried out according to the features. In practice, the conditional random field model is a undirected graph model that computes the joint probability distribution of the entire sequence of tokens given the observed sequence (words, sentence values, etc.) that needs to be signed. The conditional random field model is end-to-end, all the feature extraction work is given to the deep learning model, and X (such as X) is obtained according to the bidirectional long-time and short-time memory network₁、X₂…X_i…X_n) The possible sequence Y (e.g. Y) can be calculated using a locally adapted solution₁、Y₂…Y_i…Y_n) I.e. the final label, i.e. the named entity recognition result. Specifically, time series modeling is carried out on input text content, then a bidirectional long-time memory network is used for calculating the probability that each character takes each label, and finally a conditional random field model is used for decoding. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, when performing entity recognition on the target text to be semantically understood, performing named entity recognition on the target text to be semantically understood through a double-array dictionary tree to obtain a named entity tag of the target text to be semantically understood. Therefore, the named entity identification is carried out on the target text to be semantically understood through the double-array dictionary tree, and the named entity label of the target text to be semantically understood can be accurately identified. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the Trie is a kind of search tree, and can establish an effective data retrieval organization structure to implement an algorithm for searching words in a word bank. It is essentially a Deterministic Finite state Automaton (DFA), each node representing a state of the Automaton. Such states include "word prefix", "formed word", etc. in the dictionary. A Double Array dictionary tree (Double Array Trie) is a simple and effective realization of the Trie, and is composed of two integer arrays, wherein the index of the Array is set as i, i is an integer which is greater than or equal to 1, one Array of the Double arrays is a base value Array base [ i ], the other Array is a check Array check [ i ], and each branch of the Double arrays is a state transition from a certain state to another state after encountering a specific character. For example, for a state transition where state s encounters character c to state t, in the double array there is: check [ base [ s ] + c ] ═ s; base [ s ] + c ═ t. Before carrying out named entity recognition on a target text through a double-array dictionary tree, establishing a named entity library, and storing the named entity library by adopting the double-array dictionary tree. After the named entity library is stored by adopting the double-array dictionary tree, carrying out named entity recognition on the target text to be semantically understood through the double-array dictionary tree so as to obtain the named entity label of the target text to be semantically understood. Specifically, the target text to be semantically understood is segmented, and then the double-array dictionary tree is retrieved by using the segmentation of the target text to be semantically understood, so as to obtain the named entity label of the segmentation of the target text to be semantically understood. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, when an entity corresponding to a named entity tag matched with the entity tag configuration file in the target text to be semantically understood is replaced with a preset character, based on an entity tag matching rule in the entity tag configuration file, the named entity tag in the target text to be semantically understood is matched, and an entity corresponding to the named entity tag matched in the target text to be semantically understood is replaced with a preset character. Wherein the preset character can be a placeholder. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, when the label of the matched entity "sansheng sanshiyi peach blossom" in the target text is "tv drama", the entity "sansheng sanshiyi peach blossom" may be replaced with a placeholder "@". When the label of the entity 'Changan twelve hours' matched in the target text is 'TV play', the entity 'Changan twelve hours' can be replaced by the placeholder '@'. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, the semantic template includes a regular expression carrying the preset character. And matching the target text replaced with the placeholder based on the regular expression carrying the placeholder when performing semantic analysis on the target text replaced with the placeholder by using the semantic template associated with the entity tag configuration file so as to obtain the entity relationship to be replaced in the target text to be understood semantically. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the regular expression is a logical formula operating on a character string, and a "rule character string" is formed by using specific characters defined in advance and a combination of the specific characters, so that a text conforming to a rule is retrieved according to the "rule character string". For example, it may be determined whether a given string (e.g., a user's question) contains an entity relationship to be replaced by determining whether the user's question matches a regular expression. When the regular expression carrying the placeholder is "@ (his/her/his/her)? (daughter son child) ", if the target text that has been replaced with the placeholder matches, the entity relationship to be replaced in the target text may be" her son "," his daughter ", etc. When the regular expression carrying the placeholder is "@ (his/her/his/her)? (wife | wife) "if the target text that has been replaced with the placeholder matches, the entity relationship to be replaced in the target text may be" his wife ", etc. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S102, at least one entity corresponding to the entity relationship to be replaced is determined based on the knowledge graph.

In this embodiment, the Knowledge Graph (Knowledge Graph) is a series of different graphs displaying Knowledge development process and structural relationship in the book intelligence world, and is used for describing Knowledge resources and their carriers, mining, analyzing, constructing, drawing and displaying Knowledge and their interrelations. In particular, the knowledge-graph may be a semantic network intended to describe conceptual entities of the objective world and their relationships, sometimes referred to as a knowledge base. The knowledge-graph is composed of triplets: < entity 1, relationship, entity 2> or < entity, attribute value >, e.g., < yaoming, play-in, NBA >, < yaoming, height, 2.29m > etc. And when determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph, performing entity link on the entity relationship to be replaced by using the knowledge graph to obtain the at least one entity corresponding to the entity relationship to be replaced. The entity link is used for mapping the entity relationship to be replaced to at least one corresponding entity in the knowledge graph. Specifically, all entities possibly corresponding to the entity relationship to be replaced are searched from the knowledge graph to form a candidate entity set. And then, selecting the most possible entity from the candidate entity set as the entity corresponding to the entity relationship to be replaced. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, when a knowledge graph is used to perform entity link on the entity relationship to be replaced, the knowledge graph is called to perform entity link on the entity relationship to be replaced by using a knowledge graph proxy plug-in to obtain structured data returned by the knowledge graph; and performing data extraction on the structured data based on the storage path of the structured data to obtain at least one entity corresponding to the entity relationship to be replaced. Therefore, by using the knowledge graph agent plug-in, the appropriate knowledge graph service can be conveniently called to carry out entity link on the entity relation without extra work, and the knowledge graph service can be further made to be pluggable. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In one specific example, a Plug-in (Plug-in, also known as an add, add-in, add, or add-on, also known as an add-on) is a program written in an application program interface that conforms to a specification. It can only run under the system platform (possibly supporting multiple platforms simultaneously) specified by the program, and cannot run independently from the specified platform. Since the plug-in needs to call the function library or data provided by the original clean system. Many software has plug-ins, and there are numerous types of plug-ins. For example, in the IE, after installing the relevant plug-in, the Web browser can directly call the plug-in for processing a specific type of file. The knowledge-graph agent plug-in is a plug-in for providing knowledge-graph agent services. The structured data may be data in json format or data in xml format. The storage path of the structured data is a pre-configured data storage path. And when the structured data is subjected to data extraction based on the storage path of the structured data, extracting corresponding fields in the structured data based on the storage path of the structured data so as to obtain at least one entity corresponding to the entity relationship to be replaced. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, after obtaining at least one entity corresponding to the entity relationship to be replaced, the method further includes: and normalizing at least one entity corresponding to the entity relationship to be replaced to obtain an entity suitable for semantic understanding. Therefore, the entity suitable for semantic understanding can be obtained by normalizing at least one entity corresponding to the entity relationship to be replaced. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the normalizing at least one entity corresponding to the entity relationship to be replaced includes: unifying the case of at least one entity corresponding to the entity relationship to be replaced; or symbol screening and filtering are carried out on at least one entity corresponding to the entity relationship to be replaced; or name normalization is carried out on at least one entity corresponding to the entity relationship to be replaced. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S103, the entity relationship to be replaced in the target text to be semantically understood is replaced by the at least one entity, so as to obtain a replaced target text.

In a specific example, when the target text to be semantically understood is "play a movie of wife of zhang san", the entity relationship "wife of zhang san" to be replaced in the target text may be replaced with the entity "lie four", thereby obtaining a replaced target text "play a movie of lie four". When the target text to be semantically understood is 'playing the theme music of the peach blossom in the third generation, the entity relation' the theme music of the peach blossom in the third generation, the third generation and the tenth generation 'to be replaced in the target text can be replaced by the entity' cool, so that the replaced target text 'playing cool' is obtained. When the target text to be semantically understood is ' who the main actor of the twelve moments in Changan ' is ', the entity relationship ' the main actor of the twelve moments in Changan ' to be replaced in the target text can be replaced by the entities ' easy-to-melt Qianxiang ' and ' thunder-to-get-off ' so as to obtain the replaced target text ' who the easy-to-melt Qianxiang and the thunder-to-get-off '. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S104, performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be semantically understood.

In the embodiment of the application, the semantic understanding can be used for identifying the meaning of the text to be expressed. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, when performing semantic understanding on the replaced target text, performing semantic understanding on the replaced target text through a context-free grammar or a semantic understanding model to obtain a first semantic understanding result of the target text to be semantically understood. Therefore, the semantic understanding is carried out on the replaced target text through the context-free grammar or the semantic understanding model, and the semantic understanding result of the target text to be subjected to semantic understanding can be accurately obtained. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the context-free Grammar can be JSGF (speech Grammar Format), e.g., < fruit > < apple | pear | watermelon | mango, < fruit. For another example, < main (subject) > < i am to see < video _ type, video type > { domain) } video, intent _ video }. The semantic understanding model may be any suitable semantic understanding-capable neural network model, including but not limited to convolutional neural networks, reinforcement learning neural networks, generation networks in antagonistic neural networks, and so forth. The specific configuration of the neural network can be set by those skilled in the art according to actual requirements, such as the number of convolutional layers, the size of convolutional core, the number of channels, and so on. The semantic understanding result comprises a field, an intention and a word slot. Where a domain is a semantic understanding scene comprising a series of related intents and word slots, such as chats, weather, maps, stations, translations, stories, alarms, characters, news, music, movies, and the like. The intention is for the user to input the intended purpose, e.g., play video, play audio, etc., through interaction. The word slot is some constraint attached to the user with intent. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In one specific example, as shown in FIG. 1B, a user speaks "Play a movie of the grandma of Zhou Jewelan" to a video playback device. The voice recognition software module of the video playback device converts the user's voice into the user text "play the movie of the wife of zhou jerry". The semantic understanding pre-processing software module of the video playback device then pre-processes the user text "play the movie of the wife of zhou jersey". Specifically, semantic analysis is performed on a user text 'playing a movie of the grandma of Zhougelong' based on a semantic analysis file configured by the video service, and an entity relationship 'the grandma of Zhougelong' to be replaced in the user text 'playing a movie of the grandma of Zhougelong' is obtained. Then, using the knowledge graph, entity linking is performed on the entity relationship to be replaced, namely the wife of the shejilun, so as to obtain the entity 'kunling' corresponding to the entity relationship to be replaced, namely the wife of the shejilun. And finally, replacing the entity relationship to be replaced, namely the Mongolian grandmother, in the user text, namely the ' playing of the movie of the Mongolian grandmother ' with the corresponding entity ' Kunling ', and obtaining the replaced user text, namely ' playing of the Kunling ' movie '. Performing semantic understanding on the replaced user text 'playing the Kunzing movie' through a semantic understanding software module of the video playing device to obtain a semantic understanding result of the replaced user text 'playing the Kunzing movie', wherein the semantic understanding result comprises the fields of: video; intention is: playing the video; word slot (parameter): a movie of Kunling. The user text extracts structured information such as the field, intention and parameters of the user, the semantic elements are transmitted to a dialogue management module for inquiry processing or state management and other strategies, and finally the semantic elements are output and broadcasted to the user through voice synthesis. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

By the semantic understanding method provided by the embodiment of the invention, the semantic analysis is carried out on the target text to be semantically understood based on the configured semantic analysis file so as to obtain the entity relationship to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

The semantic understanding method provided by the present embodiment may be performed by any suitable device having data processing capabilities, including but not limited to: a camera, a terminal, a mobile terminal, a PC, a server, an in-vehicle device, an entertainment device, an advertising device, a Personal Digital Assistant (PDA), a tablet, a laptop, a handheld game machine, glasses, a watch, a wearable device, a virtual display device, a display enhancement device, or the like.

Referring to fig. 2A, a flowchart of the steps of the semantic understanding method according to the second embodiment of the present application is shown.

in step S201, based on the configured semantic parsing file, performing semantic parsing on the target text to be semantically understood to obtain an entity relationship to be replaced in the target text to be semantically understood.

Since the specific implementation of step S201 is similar to the specific implementation of step S101 in the first embodiment, it is not repeated herein.

In step S202, at least one entity corresponding to the entity relationship to be replaced is determined based on the knowledge graph.

Since the specific implementation of step S202 is similar to the specific implementation of step S102 in the first embodiment, it is not repeated herein.

In step S203, the entity relationship to be replaced in the target text to be semantically understood is replaced by the at least one entity, so as to obtain a replaced target text.

Since the specific implementation of step S203 is similar to the specific implementation of step S103 in the first embodiment, it is not repeated here.

In step S204, performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be semantically understood.

Since the specific implementation of step S204 is similar to the specific implementation of step S104 in the first embodiment, it is not repeated herein.

In step S205, performing semantic understanding on the target text to be semantically understood to obtain a second semantic understanding result of the target text to be semantically understood.

In some optional embodiments, when performing semantic understanding on the target text to be semantically understood, performing semantic understanding on the target text to be semantically understood through a context-free grammar or a semantic understanding model to obtain a second semantic understanding result of the target text to be semantically understood. Therefore, the semantic understanding is carried out on the target text to be subjected to semantic understanding through the context-free grammar or the semantic understanding model, and the semantic understanding result of the target text to be subjected to semantic understanding can be accurately obtained. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the context-free Grammar can be JSGF (speech Grammar Format), e.g., < fruit > < apple | pear | watermelon | mango, < fruit.like > < i like < fruit >. For another example, < main (subject) > < i want to see < video _ type > { domain ═ video, intent ═ play _ video }. The semantic understanding model may be any suitable semantic understanding-capable neural network model, including but not limited to convolutional neural networks, reinforcement learning neural networks, generation networks in antagonistic neural networks, and so forth. The specific configuration of the neural network can be set by those skilled in the art according to actual requirements, such as the number of convolutional layers, the size of convolutional core, the number of channels, and so on. The semantic understanding result comprises a field, an intention and a word slot. Where a domain is a semantic understanding scene comprising a series of related intents and word slots, such as chats, weather, maps, stations, translations, stories, alarms, characters, news, music, movies, and the like. The intention is for the user to input the intended purpose, e.g., play video, play audio, etc., through interaction. The word slot is some constraint attached to the user with intent. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S206, the association degrees of the first semantic understanding result and the second semantic understanding result with the business domain are determined.

In some optional embodiments, when determining the association degree between each of the first semantic understanding result and the second semantic understanding result and a business field, predicting the association degree between the first semantic understanding result and the business field by using an association degree prediction model to obtain prediction data of the association degree between the first semantic understanding result and the business field; and predicting the association degree of the second semantic understanding result and the business field through the association degree prediction model so as to obtain prediction data of the association degree of the second semantic understanding result and the business field. Therefore, the association degree of the first semantic understanding result and the association degree of the second semantic understanding result with the business field are predicted through the association degree prediction model, and the association degree of the first semantic understanding result and the association degree of the second semantic understanding result with the business field can be accurately predicted. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the prediction data of the degree of association of the first semantic understanding result with the business domain may be a prediction score of the degree of association of the first semantic understanding result with the business domain. The prediction data of the degree of association of the second semantic understanding result with the business domain may be a prediction score of the degree of association of the second semantic understanding result with the business domain. The business field can be a video field, a navigation field, a music field and the like. The degree of correlation prediction model may be any suitable neural network model that may enable feature extraction or object detection, including but not limited to convolutional neural networks, reinforcement learning neural networks, generation networks in antagonistic neural networks, and so forth. The specific configuration of the neural network can be set by those skilled in the art according to actual requirements, such as the number of convolutional layers, the size of convolutional core, the number of channels, and so on. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S207, a final semantic understanding result of the target text to be semantically understood is determined based on the association degrees between the first semantic understanding result and the second semantic understanding result and the business field, respectively.

In a specific example, when the degree of association between the first semantic understanding result and the business field is greater than the degree of association between the second semantic understanding result and the business field, determining that the final semantic understanding result of the target text to be semantically understood is the first semantic understanding result. And when the association degree of the first semantic understanding result and the service field is smaller than the association degree of the second semantic understanding result and the service field, determining that the final semantic understanding result of the target text to be semantically understood is the second semantic understanding result. When the association degree of the first semantic understanding result and the business field is equal to the association degree of the second semantic understanding result and the business field, determining that the final semantic understanding result of the target text to be semantically understood is the first semantic understanding result or the second semantic understanding result. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In one specific example, as shown in FIG. 2B, a user speaks "Play a movie of the grandma of Zhou Jewelan" to a video playback device. The voice recognition software module of the video playback device converts the user's voice into the user text "play the movie of the wife of zhou jerry". Then, the semantic understanding preprocessing software module of the video playing device preprocesses the user text "play the movie of the wife of the jersey of the week" to obtain the user text "play the movie of the kunling". Then, through a semantic understanding software module of the video playing device, semantic understanding is respectively performed on a user text "playing a movie of Kunling" and a user text "playing a movie of the grandma of Zhougelong" so as to obtain a semantic understanding result of the user text "playing a movie of Kunling" and a semantic understanding result of the user text "playing a movie of the grandma of Zhougelong". Finally, through a sequencing software module of the video playing device, the association degree of the semantic understanding result of the user text 'playing the movie of Kunling' and the association degree of the semantic understanding result of the user text 'playing the movie of the grandma of Zhonglun' with the business field 'video field' respectively are determined, and the final semantic understanding result of the user text 'playing the movie of the grandma of Zhonglun' is determined based on the association degree of the semantic understanding result of the user text 'playing the movie of Kunling' and the association degree of the semantic understanding result of the user text 'playing the movie of the grandma' with the business field 'video field'. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

On the basis of the first embodiment, performing semantic understanding on a target text to be semantically understood to obtain a second semantic understanding result of the target text to be semantically understood, determining the association degree between the first semantic understanding result and the second semantic understanding result and the business field respectively, determining a final semantic understanding result of the target text to be semantically understood based on the association degree between the first semantic understanding result and the business field respectively, and determining the final semantic understanding result of the target text to be semantically understood based on the association degree between the first semantic understanding result and the business field respectively based on the association degree between the second semantic understanding result and the business field respectively compared with other existing methods, thereby further enhancing the semantic understanding of the target text in the business field, that is, the correctness of semantic understanding of the target text in the business field can be further improved.

Referring to fig. 3, a flowchart of steps of a voice interaction method according to a third embodiment of the present application is shown.

Specifically, the voice interaction method provided by the embodiment includes the following steps:

in step S301, text recognition is performed on the received user speech data to obtain a target text to be semantically understood.

In an embodiment of the present application, the user voice data may be data of a user facing a voice application. When the text recognition is carried out on the received user voice data, the text recognition can be carried out on the received user voice data through the voice recognition model so as to obtain a target text to be semantically understood. The speech recognition model may be any suitable neural network model that can implement feature extraction or object detection, including but not limited to convolutional neural networks, reinforcement learning neural networks, generation networks in antagonistic neural networks, and so on. The specific configuration of the neural network can be set by those skilled in the art according to actual requirements, such as the number of convolutional layers, the size of convolutional core, the number of channels, and so on. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S302, based on the configured semantic parsing file, performing semantic parsing on the target text to be semantically understood to obtain an entity relationship to be replaced in the target text to be semantically understood.

Since the specific implementation of step S302 is similar to the specific implementation of step S101 in the first embodiment, it is not repeated herein.

In step S303, at least one entity corresponding to the relationship of the entity to be replaced is determined based on the knowledge graph.

Since the specific implementation of step S303 is similar to the specific implementation of step S102 in the first embodiment, the detailed description is omitted here.

In step S304, the entity relationship to be replaced in the target text to be semantically understood is replaced by the at least one entity, so as to obtain a replaced target text.

Since the specific implementation of step S304 is similar to the specific implementation of step S103 in the first embodiment, it is not repeated herein.

In step S305, performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be semantically understood.

Since the specific implementation of step S305 is similar to the specific implementation of step S104 in the first embodiment, it is not repeated herein.

In step S306, response voice data for the user voice data is generated based on the semantic understanding result.

In the embodiment of the application, the semantic understanding result comprises a field, an intention and a word slot. Where a domain is a semantic understanding scene comprising a series of related intents and word slots, such as chats, weather, maps, stations, translations, stories, alarms, characters, news, music, movies, and the like. The intention is for the user to input the intended purpose, e.g., play video, play audio, etc., through interaction. The word slot is some constraint attached to the user with intent. After obtaining the semantic understanding result, converting the semantic understanding result into response voice data by using a voice synthesis module, and feeding back the response voice data to the user. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

Through the voice interaction method provided by the embodiment of the invention, the received user voice data is subjected to text recognition to obtain a target text to be semantically understood; performing semantic analysis on the target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; and generating response voice data aiming at the user voice data based on the semantic understanding result. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

The voice interaction method provided by the present embodiment may be executed by any suitable device having data processing capability, including but not limited to: a camera, a terminal, a mobile terminal, a PC, a server, an in-vehicle device, an entertainment device, an advertising device, a Personal Digital Assistant (PDA), a tablet, a laptop, a handheld game machine, glasses, a watch, a wearable device, a virtual display device, a display enhancement device, or the like.

Referring to fig. 4, a flowchart of steps of a semantic understanding method according to the fourth embodiment of the present application is shown.

in step S401, a semantic understanding request carrying a target text to be semantically understood and sent by a terminal device is received.

In the embodiment of the present application, the terminal devices are classified into general and special types. The general-purpose terminal device generally refers to a general-purpose computer input/output device with a communication processing control function. The semantic understanding request can be understood as a request for requesting semantic understanding of the carried target text to be semantically understood. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S402, based on the configured semantic parsing file, performing semantic parsing on the target text to be semantically understood carried in the semantic understanding request to obtain an entity relationship to be replaced in the target text to be semantically understood.

Since the specific implementation of step S402 is similar to the specific implementation of step S101 in the first embodiment, it is not described herein again.

In step S403, at least one entity corresponding to the relationship of the entity to be replaced is determined based on the knowledge graph.

Since the specific implementation of step S403 is similar to the specific implementation of step S102 in the first embodiment, it is not repeated here.

In step S404, the entity relationship to be replaced in the target text to be semantically understood is replaced by the at least one entity, so as to obtain a replaced target text.

Since the specific implementation of step S404 is similar to the specific implementation of step S103 in the first embodiment, it is not repeated here.

In step S405, performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be semantically understood.

Since the specific implementation of step S405 is similar to the specific implementation of step S104 in the first embodiment, it is not described herein again.

In step S406, a semantic understanding response for the semantic understanding request is generated based on the intention information corresponding to the target text to be semantically understood.

In the embodiment of the present application, the intention information may be information of a desired purpose, such as playing video, playing music, and the like. And the semantic understanding response carries intention information corresponding to the target text to be semantically understood. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In some optional embodiments, the method further comprises: receiving a knowledge graph calling request sent by the terminal equipment; and calling the knowledge graph to provide entity link service for the terminal equipment based on the knowledge graph calling request. Therefore, the knowledge graph can be called to provide entity link service for the terminal equipment through the knowledge graph calling request. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

By the semantic understanding method provided by the embodiment of the invention, a semantic understanding request which is sent by the terminal equipment and carries a target text to be semantically understood is received; performing semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; and generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

Referring to fig. 5, a schematic structural diagram of a semantic understanding apparatus in the fifth embodiment of the present application is shown.

The semantic understanding apparatus provided in this embodiment includes: the first semantic analysis module 501 is configured to perform semantic analysis on a target text to be semantically understood based on a configured semantic analysis file to obtain an entity relationship to be replaced in the target text to be semantically understood; a first determining module 502, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced; a first replacing module 503, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; a first semantic understanding module 504, configured to perform semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be semantically understood.

The semantic understanding apparatus provided in this embodiment is used to implement the corresponding semantic understanding method in the foregoing method embodiments, and has the beneficial effects of the corresponding method embodiments, which are not described herein again.

Referring to fig. 6, a schematic structural diagram of a semantic understanding apparatus in the sixth embodiment of the present application is shown.

The semantic understanding apparatus provided in this embodiment includes: the first semantic parsing module 601 is configured to perform semantic parsing on a target text to be semantically understood based on a configured semantic parsing file to obtain an entity relationship to be replaced in the target text to be semantically understood; a first determining module 602, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced; a first replacing module 604, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; a first semantic understanding module 605, configured to perform semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be semantically understood.

Optionally, the semantic parsing file includes an entity tag configuration file and a semantic template associated with the entity tag configuration file, and the first semantic parsing module 601 includes: an entity identifying unit 6011, configured to perform named entity identification on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood; a replacing unit 6012, configured to replace, in the target text to be semantically understood, an entity corresponding to a named entity tag matched with the entity tag configuration file with a preset character; a semantic parsing unit 6013, configured to perform semantic parsing on the target text that has been replaced with the preset character by using a semantic template associated with the entity tag configuration file, so as to obtain an entity relationship to be replaced in the target text to be understood semantically.

Optionally, the entity identifying unit 6011 is specifically configured to: and carrying out named entity recognition on the target text to be semantically understood through a named entity recognition model so as to obtain a named entity label of the target text to be semantically understood.

Optionally, the entity identifying unit 6011 is specifically configured to: and carrying out named entity recognition on the target text to be semantically understood through a double-array dictionary tree to obtain a named entity label of the target text to be semantically understood.

Optionally, the semantic template includes a regular expression carrying the preset character.

Optionally, the first determining module 602 includes: an entity linking unit 6021, configured to perform entity linking on the entity relationship to be replaced by using the knowledge graph to obtain at least one entity corresponding to the entity relationship to be replaced, where the entity linking is used to map the entity relationship to be replaced to the corresponding at least one entity in the knowledge graph.

Optionally, the entity linking unit 6021 is specifically configured to: calling the knowledge graph to perform entity link on the entity relationship to be replaced by using a knowledge graph proxy plug-in so as to obtain structured data returned by the knowledge graph; and performing data extraction on the structured data based on the storage path of the structured data to obtain at least one entity corresponding to the entity relationship to be replaced.

Optionally, after the first determining module 602, the apparatus further includes: a normalization module 603, configured to normalize at least one entity corresponding to the entity relationship to be replaced, so as to obtain an entity suitable for semantic understanding.

Optionally, the first semantic understanding module 605 is specifically configured to: and performing semantic understanding on the replaced target text through a context-free grammar or a semantic understanding model to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

Optionally, after the first semantic understanding module 605, the apparatus further includes: a second semantic understanding module 606, configured to perform semantic understanding on the target text to be subjected to semantic understanding to obtain a second semantic understanding result of the target text to be subjected to semantic understanding; a second determining module 607, configured to determine the association degrees of the first semantic understanding result and the second semantic understanding result with the business domain respectively; a third determining module 608, configured to determine a final semantic understanding result of the target text to be semantically understood based on the association degrees between the first semantic understanding result and the second semantic understanding result, and the business field, respectively.

Referring to fig. 7, a schematic structural diagram of a speech interaction apparatus in the seventh embodiment of the present application is shown.

The voice interaction device provided by the embodiment comprises: a text recognition module 701, configured to perform text recognition on received user voice data to obtain a target text to be semantically understood; a second semantic parsing module 702, configured to perform semantic parsing on the target text to be semantically understood based on the configured semantic parsing file, so as to obtain an entity relationship to be replaced in the target text to be semantically understood; a fourth determining module 703, configured to determine, based on the knowledge graph, at least one entity corresponding to the entity relationship to be replaced; a second replacing module 704, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; a third semantic understanding module 705, configured to perform semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; a first generating module 706, configured to generate response voice data for the user voice data based on the semantic understanding result.

The voice interaction apparatus provided in this embodiment is used to implement the corresponding voice interaction method in the foregoing multiple method embodiments, and has the beneficial effects of the corresponding method embodiments, which are not described herein again.

Referring to fig. 8, a schematic structural diagram of a semantic understanding apparatus in the eighth embodiment of the present application is shown.

The semantic understanding apparatus provided in this embodiment includes: a first receiving module 801, configured to receive a semantic understanding request that is sent by a terminal device and carries a target text to be semantically understood; a third semantic parsing module 802, configured to perform semantic parsing on a target text to be semantically understood, which is carried in the semantic understanding request, based on a configured semantic parsing file, so as to obtain an entity relationship to be replaced in the target text to be semantically understood; a fifth determining module 803, configured to determine, based on the knowledge graph, at least one entity corresponding to the entity relationship to be replaced; a third replacing module 804, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text; a fourth semantic understanding module 805, configured to perform semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; a second generating module 806, configured to generate a semantic understanding response for the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

Optionally, the apparatus further comprises: a second receiving module 807, configured to receive a knowledge graph invoking request sent by the terminal device; a calling module 808, configured to call the knowledge graph to provide an entity link service to the terminal device based on the knowledge graph call request.

Referring to fig. 9, a schematic structural diagram of an electronic device according to a ninth embodiment of the present invention is shown, and the specific embodiment of the present invention does not limit the specific implementation of the electronic device.

As shown in fig. 9, the electronic device may include: a processor (processor)902, a communication Interface 904, a memory 906, and a communication bus 908.

Wherein:

the processor 902, communication interface 904, and memory 906 communicate with one another via a communication bus 908.

A communication interface 904 for communicating with other electronic devices or servers.

The processor 902, configured to execute the program 910, may specifically perform relevant steps in the above semantic understanding method embodiment or the voice interaction method embodiment.

In particular, the program 910 may include program code that includes computer operating instructions.

The processor 902 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement an embodiment of the invention. The intelligent device comprises one or more processors which can be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

A memory 906 for storing a program 910. The memory 906 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 910 may specifically be configured to cause the processor 902 to perform the following operations: performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

In an alternative embodiment, the semantic parsing file includes an entity tag profile and a semantic template associated with the entity tag profile. The program 910 is further configured to enable the processor 902, when performing semantic parsing on a target text to be semantically understood based on the configured semantic parsing file, perform named entity identification on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood; replacing an entity corresponding to the named entity tag matched with the entity tag configuration file in the target text to be semantically understood with a preset character; and performing semantic analysis on the target text which is replaced by the preset characters by using a semantic template associated with the entity tag configuration file to obtain the entity relationship to be replaced in the target text to be understood semantically.

In an optional implementation manner, the program 910 is further configured to, when performing named entity recognition on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood, perform named entity recognition on the target text to be semantically understood by using a named entity recognition model to obtain the named entity tag of the target text to be semantically understood.

In an optional implementation manner, the program 910 is further configured to enable the processor 902, when performing named entity recognition on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood, perform named entity recognition on the target text to be semantically understood through a double-array dictionary tree to obtain the named entity tag of the target text to be semantically understood.

In an optional implementation manner, the semantic template includes a regular expression carrying the preset character.

In an optional implementation manner, the program 910 is further configured to, when determining, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced, use the knowledge graph to perform entity linking on the entity relationship to be replaced, so as to obtain the at least one entity corresponding to the entity relationship to be replaced, where the entity linking is used to map the entity relationship to be replaced to the corresponding at least one entity in the knowledge graph.

In an optional implementation manner, the program 910 is further configured to enable the processor 902, when using the knowledge graph to perform entity linking on the entity relationship to be replaced so as to obtain at least one entity corresponding to the entity relationship to be replaced, call the knowledge graph to perform entity linking on the entity relationship to be replaced by using a knowledge graph proxy plugin so as to obtain structured data returned by the knowledge graph; and performing data extraction on the structured data based on the storage path of the structured data to obtain at least one entity corresponding to the entity relationship to be replaced.

In an optional implementation manner, the program 910 is further configured to enable the processor 902, after obtaining the at least one entity corresponding to the entity relationship to be replaced, to normalize the at least one entity corresponding to the entity relationship to be replaced, so as to obtain an entity suitable for semantic understanding.

In an optional implementation, the program 910 is further configured to, when performing semantic understanding on the replaced target text, perform semantic understanding on the replaced target text through a context-free grammar or a semantic understanding model, so as to obtain a first semantic understanding result of the target text to be semantically understood.

In an optional implementation manner, the program 910 is further configured to enable the processor 902, after performing semantic understanding on the replaced target text, to perform semantic understanding on the target text to be subjected to semantic understanding to obtain a second semantic understanding result of the target text to be subjected to semantic understanding; determining the degree of association between the first semantic understanding result and the second semantic understanding result and the business field respectively; and determining a final semantic understanding result of the target text to be semantically understood based on the association degree between the first semantic understanding result and the second semantic understanding result and the business field respectively.

For specific implementation of each step in the program 910, reference may be made to corresponding steps and corresponding descriptions in units in the foregoing semantic understanding method embodiments, which are not described herein again. It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described devices and modules may refer to the corresponding process descriptions in the foregoing method embodiments, and are not described herein again.

Through the electronic equipment of the embodiment, semantic analysis is performed on a target text to be semantically understood based on a configured semantic analysis file so as to obtain an entity relationship to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity to obtain a replaced target text; and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

The program 910 may specifically be configured to cause the processor 902 to perform the following operations: performing text recognition on received user voice data to obtain a target text to be semantically understood; performing semantic analysis on the target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; and generating response voice data aiming at the user voice data based on the semantic understanding result.

For specific implementation of each step in the program 910, reference may be made to corresponding steps and corresponding descriptions in units in the foregoing voice interaction method embodiments, which are not described herein again. It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described devices and modules may refer to the corresponding process descriptions in the foregoing method embodiments, and are not described herein again.

Through the electronic equipment of the embodiment, text recognition is carried out on the received user voice data to obtain a target text to be semantically understood; performing semantic analysis on the target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding; and generating response voice data aiming at the user voice data based on the semantic understanding result. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

The program 910 may specifically be configured to cause the processor 902 to perform the following operations: receiving a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood; performing semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; and generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

Through the electronic equipment of the embodiment, a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood is received; performing semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood; determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph; replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text; performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding; and generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood. Compared with the existing other modes, the semantic understanding of the target text in the business field can be effectively enhanced by replacing the entity relationship to be replaced in the target text to be semantically understood with at least one entity corresponding to the entity relationship to be replaced and performing semantic understanding on the target text of the sentence after replacement.

It should be noted that, according to the implementation requirement, each component/step described in the embodiment of the present invention may be divided into more components/steps, and two or more components/steps or partial operations of the components/steps may also be combined into a new component/step to achieve the purpose of the embodiment of the present invention.

The above-described method according to an embodiment of the present invention may be implemented in hardware, firmware, or as software or computer code storable in a recording medium such as a CD ROM, a RAM, a floppy disk, a hard disk, or a magneto-optical disk, or as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium downloaded through a network and to be stored in a local recording medium, so that the method described herein may be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware such as an ASIC or FPGA. It is understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code that, when accessed and executed by a computer, processor, or hardware, implements the semantic understanding method or voice interaction method described herein. Further, when a general-purpose computer accesses code for implementing the semantic understanding method or the voice interaction method illustrated herein, execution of the code transforms the general-purpose computer into a special-purpose computer for performing the semantic understanding method or the voice interaction method illustrated herein.

Those of ordinary skill in the art will appreciate that the various illustrative elements and method steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present embodiments.

The above embodiments are only for illustrating the embodiments of the present invention and not for limiting the embodiments of the present invention, and those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention, so that all equivalent technical solutions also belong to the scope of the embodiments of the present invention, and the scope of patent protection of the embodiments of the present invention should be defined by the claims.

Claims

1. A method of semantic understanding, the method comprising:

performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood;

determining at least one entity corresponding to the entity relationship to be replaced based on a knowledge graph;

replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity to obtain a replaced target text;

and performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

2. The method of claim 1, wherein the semantic parsing file comprises an entity tag profile and a semantic template associated with the entity tag profile,

the semantic parsing of the target text to be semantically understood based on the configured semantic parsing file comprises the following steps:

carrying out named entity recognition on the target text to be semantically understood to obtain a named entity label of the target text to be semantically understood;

replacing an entity corresponding to the named entity tag matched with the entity tag configuration file in the target text to be semantically understood with a preset character;

and performing semantic analysis on the target text which is replaced by the preset characters by using a semantic template associated with the entity tag configuration file to obtain the entity relationship to be replaced in the target text to be understood semantically.

3. The method according to claim 2, wherein the performing named entity recognition on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood comprises:

and carrying out named entity recognition on the target text to be semantically understood through a named entity recognition model so as to obtain a named entity label of the target text to be semantically understood.

4. The method according to claim 2, wherein the performing named entity recognition on the target text to be semantically understood to obtain a named entity tag of the target text to be semantically understood comprises:

and carrying out named entity recognition on the target text to be semantically understood through a double-array dictionary tree to obtain a named entity label of the target text to be semantically understood.

5. The method of claim 2, wherein the semantic template comprises a regular expression carrying the preset characters.

6. The method of claim 1, wherein the determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge-graph comprises:

and using the knowledge graph to perform entity linkage on the entity relationship to be replaced so as to obtain at least one entity corresponding to the entity relationship to be replaced, wherein the entity linkage is used for mapping the entity relationship to be replaced to the corresponding at least one entity in the knowledge graph.

7. The method of claim 6, wherein the using the knowledge-graph to perform entity linking on the entity relationship to be replaced to obtain at least one entity corresponding to the entity relationship to be replaced comprises:

calling the knowledge graph to perform entity link on the entity relationship to be replaced by using a knowledge graph proxy plug-in so as to obtain structured data returned by the knowledge graph;

and performing data extraction on the structured data based on the storage path of the structured data to obtain at least one entity corresponding to the entity relationship to be replaced.

8. The method according to claim 1, wherein after determining at least one entity corresponding to the entity relationship to be replaced, the method further comprises:

and normalizing at least one entity corresponding to the entity relationship to be replaced to obtain an entity suitable for semantic understanding.

9. The method of claim 1, wherein the semantically understanding the replaced target text comprises:

and performing semantic understanding on the replaced target text through a context-free grammar or a semantic understanding model to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

10. The method of claim 1, wherein after semantically understanding the replaced target text, the method further comprises:

performing semantic understanding on the target text to be subjected to semantic understanding to obtain a second semantic understanding result of the target text to be subjected to semantic understanding;

determining the degree of association between the first semantic understanding result and the second semantic understanding result and the business field respectively;

and determining a final semantic understanding result of the target text to be semantically understood based on the association degree between the first semantic understanding result and the second semantic understanding result and the business field respectively.

11. A method of voice interaction, the method comprising:

performing text recognition on received user voice data to obtain a target text to be semantically understood;

performing semantic analysis on the target text to be semantically understood based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood;

performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding;

and generating response voice data aiming at the user voice data based on the semantic understanding result.

12. A method of semantic understanding, the method comprising:

receiving a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood;

performing semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file to obtain an entity relation to be replaced in the target text to be semantically understood;

performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding;

and generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

13. The method of claim 12, wherein the method further comprises:

receiving a knowledge graph calling request sent by the terminal equipment;

and calling the knowledge graph to provide entity link service for the terminal equipment based on the knowledge graph calling request.

14. A semantic understanding apparatus, the apparatus comprising:

the first semantic analysis module is used for performing semantic analysis on a target text to be semantically understood based on the configured semantic analysis file so as to obtain an entity relation to be replaced in the target text to be semantically understood;

the first determination module is used for determining at least one entity corresponding to the entity relationship to be replaced based on the knowledge graph;

a first replacement module, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text;

and the first semantic understanding module is used for performing semantic understanding on the replaced target text to obtain a first semantic understanding result of the target text to be subjected to semantic understanding.

15. A voice interaction apparatus, the apparatus comprising:

the text recognition module is used for performing text recognition on the received user voice data to obtain a target text to be semantically understood;

the second semantic analysis module is used for carrying out semantic analysis on the target text to be semantically understood based on the configured semantic analysis file so as to obtain the entity relationship to be replaced in the target text to be semantically understood;

a fourth determining module, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced;

the second replacement module is used for replacing the entity relationship to be replaced in the target text to be semantically understood with the at least one entity so as to obtain a replaced target text;

the third semantic understanding module is used for performing semantic understanding on the replaced target text to obtain a semantic understanding result of the target text to be subjected to semantic understanding;

and the first generation module is used for generating response voice data aiming at the user voice data based on the semantic understanding result.

16. A semantic understanding apparatus, the apparatus comprising:

the first receiving module is used for receiving a semantic understanding request which is sent by terminal equipment and carries a target text to be semantically understood;

the third semantic analysis module is used for carrying out semantic analysis on a target text to be semantically understood carried in the semantic understanding request based on the configured semantic analysis file so as to obtain an entity relation to be replaced in the target text to be semantically understood;

a fifth determining module, configured to determine, based on a knowledge graph, at least one entity corresponding to the entity relationship to be replaced;

a third replacing module, configured to replace the entity relationship to be replaced in the target text to be semantically understood with the at least one entity, so as to obtain a replaced target text;

the fourth semantic understanding module is used for performing semantic understanding on the replaced target text to obtain intention information corresponding to the target text to be subjected to semantic understanding;

and the second generation module is used for generating a semantic understanding response aiming at the semantic understanding request based on the intention information corresponding to the target text to be semantically understood.

17. An electronic device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, which causes the processor to execute the operation corresponding to the semantic understanding method according to any one of claims 1 to 10, or execute the operation corresponding to the voice interaction method according to claim 11, or execute the operation corresponding to the semantic understanding method according to claim 12 or 13.

18. A computer storage medium having stored thereon a computer program which, when executed by a processor, implements a semantic understanding method according to any one of claims 1 to 10, or performs a voice interaction method according to claim 11, or performs a semantic understanding method according to claim 12 or 13.