Data processing method, device, electronic device and storage medium

Abstract

A data processing method, device, electronic device and storage medium. The method comprises: obtaining audio data (301); recognizing the audio data to obtain a recognized text; extracting a target word in the recognized text; and extracting a first word and a second word of the target word in the recognized text (302); determining a first word vector of the target word; and determining at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector and at least one of the second word vector and the third word vector, searching a vector set for first information associated with the target word (303); generating a simultaneous interpretation result using the first information and the recognized text; and outputting the simultaneous interpretation result (304).

Description

Technical Field

[0001] The present application relates to simultaneous interpretation technology, and in particular to a data processing method, device, electronic device and storage medium. Background Art

[0002] Machine simultaneous interpretation technology is a voice translation product that has emerged in recent years for scenarios such as meetings and reports. It combines automatic speech recognition technology (ASR, Automatic Speech Recognition) technology and machine translation (MT, Machine Translation) technology to provide multi-language subtitles for the speaker's speech content, replacing manual simultaneous interpretation services.

[0003] In related machine simultaneous interpretation technologies, the speech content is usually translated and displayed through text, but the displayed content cannot enable users to fully and accurately understand the speech content. Summary of the invention

[0004] Embodiments of the present application provide a data processing method, device, electronic device, and storage medium.

[0005] The present application provides a data processing method, including:

[0006] Get audio data;

[0007] Recognize the audio data to obtain a recognized text; extract a target word from the recognized text; and extract a first word and a second word of the target word from the recognized text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word;

[0008] Determine a first word vector of the target word; and determine at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, search for first information associated with the target word from a vector set; the first information is used to interpret the meaning of the target word;

[0009] Generate a simultaneous interpretation result using the first information and the recognized text; output the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data;

[0010] Among them, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node that has an associated relationship with the first knowledge node.

[0011] The present application provides a data processing device, including:

[0012] An acquisition unit, used for acquiring audio data;

[0013] A first processing unit is configured to recognize the audio data to obtain a recognized text; extract a target word from the recognized text; and extract a first word and a second word of the target word from the recognized text; the first word is a word in the previous context of the target word; and the second word is a word in the following context of the target word;

[0014] A second processing unit is used to determine a first word vector of the target word; and determine at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, search for first information associated with the target word from a vector set; the first information is used to interpret the meaning of the target word;

[0015] A third processing unit is used to generate a simultaneous interpretation result using the first information and the recognized text; and output the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data;

[0016] Among them, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node that has an associated relationship with the first knowledge node.

[0017] An embodiment of the present application further provides an electronic device, comprising: a processor and a memory for storing a computer program that can be run on the processor;

[0018] Wherein, the processor is used to execute the steps of any of the above methods when running the computer program.

[0019] An embodiment of the present application further provides a storage medium, wherein the medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0020] The data processing method, device, electronic device and storage medium provided in the embodiments of the present application obtain audio data; recognize the audio data to obtain a recognized text; extract a target word in the recognized text; and extract a first word and a second word of the target word in the recognized text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word; determine a first word vector of the target word; and determine at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, search a vector set for first information associated with the target word; the first information is used to interpret the meaning of the target word; generate a simultaneous interpretation result using the first information and the recognized text; output the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data; wherein the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy an embedded expression; the at least two entity vectors represent attribute information of a first knowledge node in a knowledge graph database and attribute information of at least one second knowledge node associated with the first knowledge node. By adopting the technical solution of the embodiment of the present application, the first information associated with the target word can be searched from the vector set based on the first word vector, and at least one of the second word vector and the third word vector. Compared with the method of searching for the explanation information of the target word from the knowledge graph database in the related art, there is no need to perform the disambiguation step, the translation accuracy is high, and the implementation method is simple. In this way, it can help users fully and accurately understand the content of the speaker's speech and reduce the difficulty of users' understanding of the content of the speech. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of determining the meaning of a proper noun in a speech content in the related art;

[0022] Figure 2 Schematic diagram of the architecture of the machine simultaneous translation system in the related technology

[0023] Figure 3 It is a schematic diagram of the implementation flow of the data processing method of the embodiment of the present application;

[0024] Figure 4 It is a schematic diagram of an implementation flow of determining a target word in a recognized text corresponding to audio data in an embodiment of the present application;

[0025] Figure 5 This is a schematic diagram of a first implementation flow of searching a vector set for first information associated with a target word in an embodiment of the present application;

[0026] Figure 6This is a first schematic diagram of the first information of the embodiment of the present application;

[0027] Figure 7 This is a schematic diagram of a second implementation flow of searching for first information associated with a target word from a vector set in an embodiment of the present application;

[0028] Figure 8 It is a second schematic diagram of the first information of the embodiment of the present application;

[0029] Fig. 9 This is a schematic diagram of a third implementation flow of searching for first information associated with a target word from a vector set in an embodiment of the present application;

[0030] Fig.10 This is a third schematic diagram of the first information of the embodiment of the present application;

[0031] Fig.11 It is a schematic diagram of outputting simultaneous interpretation results in an embodiment of the present application;

[0032] Fig.12 It is a schematic diagram of the composition structure of the data processing device according to an embodiment of the present application;

[0033] Fig.13 It is a schematic diagram of the composition structure of the electronic device according to the embodiment of the present application. DETAILED DESCRIPTION

[0034] Before describing the technical solutions of the embodiments of the present application in detail, a brief description of the related technologies is first given.

[0035] In the related art, in the simultaneous interpretation scenario, in order to translate the proper nouns in the speaker's speech content, the first method is to collect words in multiple technical fields, use the collected words to configure a terminology dictionary, and the terminology dictionary stores the correspondence between words and key values; combined with the constructed AC state machine, the key value matching the proper noun in the speech content is searched in the terminology dictionary, and the interpretation information of the words corresponding to the key value is used as the interpretation information of the proper noun in the speech content. The second method is to search for the interpretation information of the proper noun in the speech content from the knowledge graph database. The defect of the first method is that the proper noun in the speech content may correspond to multiple key values, so the interpretation information of the proper noun in the speech content retrieved from the terminology dictionary may contain multiple meanings, resulting in inaccurate translation results. The defect of the second method is that the proper noun in the speech content may have multiple meanings, so there are more than one knowledge nodes associated with the proper noun in the speech content retrieved from the knowledge graph database, and the interpretation information of the proper noun needs to be obtained by executing the disambiguation step, which is more complicated to implement. Figure 1 It is a schematic diagram of determining the meaning of proper nouns in speech content in the related art, such as Figure 1 As shown, for the proper noun "apple", two interpretation information with different meanings can be obtained.

[0036] Figure 2 Schematic diagram of the architecture of a machine simultaneous translation system in the related art; Figure 2 As shown, the system may include: a machine simultaneous interpretation server, a speech recognition server, a translation server, a mobile terminal sending server, an audience mobile terminal, a personal computer (PC) client, and a display screen.

[0037] In actual applications, a speaker can give a conference speech through a PC client. During the conference speech, the PC client collects the speaker's audio and sends the collected audio to the machine simultaneous interpretation server. The machine simultaneous interpretation server recognizes the audio data through a speech recognition server to obtain a recognized text, and then translates the recognized text through a translation server to obtain a translation result; the machine simultaneous interpretation server sends the translation result to the PC client, and sends the translation result to the audience's mobile terminal through a mobile terminal sending server, and displays the translation result to the user, thereby translating the speaker's speech content into the language required by the user and displaying it.

[0038] Based on this, in an embodiment of the present application, audio data is obtained; the audio data is recognized to obtain a recognized text; a target word in the recognized text is extracted; and a first word and a second word of the target word in the recognized text are extracted; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word; a first word vector of the target word is determined; and a second word vector of the first word and a third word vector of the second word are determined; based on the first word vector, and at least one of the second word vector and the third word vector, first information associated with the target word is searched from a vector set; the first information is used to interpret the meaning of the target word; a simultaneous interpretation result is generated using the first information and the recognized text; the simultaneous interpretation result is output; the simultaneous interpretation result is used to be presented on a display device when the audio data is played; wherein the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy an embedded expression; the at least two entity vectors represent attribute information of a first knowledge node in a knowledge graph database and attribute information of at least one second knowledge node having an associated relationship with the first knowledge node.

[0039] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] The present application embodiment provides a data processing method, Figure 3is a schematic diagram of the implementation flow of the data processing method of the embodiment of the present application, such as Figure 3 As shown, the following steps are included:

[0041] Step 301: Acquire audio data;

[0042] Step 302: Recognize the audio data to obtain a recognized text; extract the target word in the recognized text; and extract the first word and the second word of the target word in the recognized text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word;

[0043] Step 303: determining a first word vector of the target word; and determining at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, searching for first information associated with the target word from a vector set;

[0044] The first information is used to explain the meaning of the target word;

[0045] Step 304: Generate a simultaneous interpretation result using the first information and the recognized text; output the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data;

[0046] Here, in step 301, in actual application, the audio data may be the audio data of a speaker's speech in a conference scenario of a simultaneous interpretation application.

[0047] Here, in step 302, in actual application, the target word may be a proper noun in the content of the audio data, and specifically may refer to a word with multiple meanings, for example, "Apple|" may refer to apple among fruits, or may refer to an iPhone.

[0048] Here, in step 303, in actual application, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node that has an associated relationship with the first knowledge node.

[0049] Here, in step 304, in actual application, the display device may specifically be a client.

[0050] In actual application, the data processing method can be applied to a server or a terminal. When the data processing method is applied to a server, the terminal can collect the audio data, and then the server can obtain the audio data from the terminal.

[0051] Here, the present application does not limit the specific type of the terminal, for example, it can be a smart phone, a personal computer, a laptop computer, a tablet computer, and a portable wearable device.

[0052] Here, in actual application, the associated first information contained in the content of the audio data may be determined in the following manner.

[0053] Specifically, the content of the audio data may involve words with multiple meanings. If the interpretation information of the word is searched directly from the knowledge graph database, the interpretation information corresponding to multiple knowledge nodes may be searched, resulting in the interpretation information of the word being non-unique, and it is necessary to filter it through the disambiguation step, which is more complicated to implement. In order to be able to retrieve the unique interpretation information of the word, a vector set can be constructed using the knowledge nodes in the knowledge graph database. In this way, the first information used to explain the meaning of the word can be found from the vector set. Compared with the method of directly searching for the interpretation information of the word from the knowledge graph database in the related art, there is no need to perform the disambiguation step, and the unique interpretation information of the word can be accurately determined, which significantly improves the accuracy of the retrieval.

[0054] In actual application, in a conference scenario where simultaneous interpretation is used, the content of the speech by the conference speaker may contain one or more technical terms in the technical field. In order to help the recipient of the audio data accurately understand the content of the conference speaker's speech, the proper nouns contained in the content of the audio data can be searched from the dictionary database.

[0055] Based on this, in one embodiment, the step of extracting the target words from the recognized text includes:

[0056] extracting at least two keywords from the recognized text;

[0057] For each of the at least two keywords, perform entity recognition on the corresponding keyword to obtain at least two entity terms;

[0058] Searching for proper nouns matching the at least two entity words from a dictionary database;

[0059] The entity words corresponding to the found proper nouns are used as the target words.

[0060] Among them, entity recognition refers to identifying entity words with specific meanings in text, such as names of people, places, time, organization names, etc. Proper nouns can refer to words with multiple meanings.

[0061] Here, at least two entity phrases with specific meanings may also be extracted, which is not limited here.

[0062] Specifically, keyword extraction technology can be used to extract keywords from the recognition text to obtain at least two keywords; the recognition text can also be converted into multiple bit data; bit data whose bit number meets a preset threshold is searched from the multiple bit data; and the bit data that meets the preset conditions is used as the keyword.

[0063] The specific situation of performing entity recognition on the at least two keywords is described below.

[0064] In the first case, entity recognition is performed on the at least two keywords using preset rules to obtain at least two entity words.

[0065] For example, a keyword matching a preset character string is searched from the at least two keywords using a regular expression, and the searched keyword matching the preset character string is used as an entity word.

[0066] In the second case, a classification model is used to perform entity recognition on the at least two keywords to obtain at least two entity words with specific meanings.

[0067] For example, a neural network model is used to map the at least two keywords from input to output to obtain a recognition result; when the recognition result indicates that the corresponding keyword has a specific meaning, the corresponding keyword is used as an entity word.

[0068] In a third case, a sequence labeling model is used to perform entity recognition on the at least two keywords to obtain at least two entity words.

[0069] For example, the semantics of the at least one keyword is analyzed using semantic analysis technology to obtain semantic information of related keywords, the parts of speech of the at least two keywords are marked according to the obtained semantic information, and keywords with specific meanings are selected from the marked keywords as entity words.

[0070] Here, proper nouns in the content of audio data are determined by entity recognition technology and dictionary database, which has the following advantages:

[0071] If proper nouns contained in the content of the audio data are found in the dictionary database, a vector set can be subsequently constructed using the knowledge nodes in the knowledge graph database, that is, the attributes of the knowledge nodes in the knowledge graph database are represented by vectors, and the vectors corresponding to at least two knowledge nodes with an associated relationship are operated through embedded expressions to obtain a column vector or a row vector, and a vector set is obtained based on the obtained column vector or row vector. In this way, the context of the found proper noun can be used to search for first information for explaining the meaning of the proper noun from the vector set. Compared with the method of directly searching for the explanation information of the proper noun from the knowledge graph database in the related art, there is no need to perform the disambiguation step, and the explanation information of the proper noun can be accurately determined.

[0072] In one example, taking the simultaneous interpretation scenario as an example, a schematic diagram of the implementation process of determining the target words in the recognized text corresponding to the audio data is described, such as Figure 4 As shown, including:

[0073] Step 1: extract at least two keywords from the recognition text corresponding to the audio data;

[0074] Assume that the recognized text corresponding to the audio data is "Among many mobile phone brands, I like Apple the most."

[0075] The at least two keywords may be “brand, like, apple”;

[0076] Step 2: for each of the at least two keywords, perform entity recognition on the corresponding keyword to obtain at least two entity terms;

[0077] The at least two entity words may be “brand, apple”.

[0078] Step 3: searching a dictionary database for proper nouns matching the at least two entity words;

[0079] No proper noun matching “brand” is found in the dictionary database; the proper noun matching “apple” found in the dictionary database is “apple”.

[0080] Step 4: Use the entity words corresponding to the found proper nouns as the target words.

[0081] The target word included in the content of the audio data is "apple".

[0082] In this example, in the scenario of speech recognition, the speech data input by the user is obtained, the speech data is recognized to obtain a recognized text, and a search is made from a dictionary database to determine whether the recognized text contains professional terms with multiple meanings. If the dictionary database finds that the recognized text contains proper nouns, the vector set can be used to directly determine the unique interpretation information of the proper nouns, thereby helping the user to accurately understand the content of the audio data and improve the user experience.

[0083] Here, in actual application, the interpretation information of the proper nouns in the content of the audio data can be determined from the vector set in the following three ways.

[0084] In a first manner, the interpretation information of the proper nouns in the content of the audio data is determined from a vector set by using the proper nouns in the content of the audio data and the words in the context of the proper nouns.

[0085] Specifically, a vector set is constructed using knowledge nodes in the knowledge graph database, that is, the attributes of the knowledge nodes in the knowledge graph database are represented by vectors, and the vectors corresponding to two knowledge nodes with an associated relationship are operated through embedded expressions to obtain a column vector or a row vector, and a vector set is obtained based on the obtained column vector or row vector. In this way, the proper nouns in the content of the audio data and the words in the context of the proper nouns can be used to find the corresponding column vector or row vector from the vector set; based on the found column vector or row vector, the knowledge information of the corresponding knowledge node is obtained to obtain the interpretation information of the proper nouns in the content of the audio data.

[0086] Each knowledge node in the knowledge graph database can be represented by a triple, for example, (a, b, c) is used to represent a knowledge node, where a represents the identifier of the knowledge node in the knowledge graph database; b represents the number of other knowledge nodes associated with the knowledge node; and c represents the identifiers of other knowledge nodes associated with the knowledge node. Other knowledge nodes associated with the knowledge node can be determined from the two dimensions of the depth and breadth of the knowledge information of the knowledge node.

[0087] The second way is to determine the interpretation information of the proper nouns in the content of the audio data from the vector set by using the proper nouns in the content of the audio data and the words in the context of the proper nouns.

[0088] Specifically, a vector set is constructed using knowledge nodes in the knowledge graph database, that is, the attributes of the knowledge nodes in the knowledge graph database are represented by vectors, and the vectors corresponding to two knowledge nodes with an associated relationship are operated through embedded expressions to obtain a column vector or a row vector, and a vector set is obtained based on the obtained column vector or row vector. In this way, the proper nouns in the content of the audio data and the words in the context of the proper nouns can be used to find the corresponding column vector or row vector from the vector set; based on the found column vector or row vector, the knowledge information of the corresponding knowledge node is obtained to obtain the interpretation information of the proper nouns in the content of the audio data.

[0089] A third method is to determine the interpretation information of the proper nouns in the content of the audio data from a vector set by using the proper nouns in the content of the audio data, the words in the preceding context of the proper nouns, and the words in the following context of the proper nouns.

[0090] Specifically, a vector set is constructed using knowledge nodes in the knowledge graph database, that is, the attributes of the knowledge nodes in the knowledge graph database are represented by vectors, and the vectors corresponding to three knowledge nodes with an associated relationship are operated through embedded expressions to obtain a column vector or a row vector, and a vector set is obtained based on the obtained column vector or row vector. In this way, the proper nouns in the content of the audio data, the words in the context of the proper nouns, and the words in the context of the proper nouns can be used to find the corresponding column vector or row vector from the vector set; based on the found column vector or row vector, the knowledge information of the corresponding knowledge node is obtained to obtain the interpretation information of the proper nouns in the content of the audio data.

[0091] In actual application, through embedded expressions, the proper nouns in the content of the audio data, the words in the context of the proper nouns, and the words in the context of the proper nouns can be represented by a vector, so that the first information associated with the proper nouns can be found from the vector set.

[0092] Based on this, in one embodiment, searching for first information associated with the target word from a vector set based on the first word vector and at least one of the second word vector and the third word vector includes:

[0093] Combining the first word vector with the fourth word vector to obtain at least one vector group; the fourth word vector is at least one of the second word vector and the third word vector;

[0094] Finding a first vector matching the at least one vector group from the vector set;

[0095] Based on the at least two entity vectors corresponding to the first vector, searching for at least two knowledge nodes corresponding to the at least two entity vectors from a knowledge graph database; the knowledge graph database stores the relationship between the knowledge nodes and the entity vectors;

[0096] Acquire knowledge information of the at least two knowledge nodes; and use the acquired knowledge information of the at least two knowledge nodes as first information associated with the target word.

[0097] Specifically, the first word vector and the fourth word vector are combined to obtain at least one vector group, including the following combinations:

[0098] The first method is to combine the first word vector and the second word vector to obtain a vector group;

[0099] The first word vector and the second word vector can be combined by the embedded expression shown in formula (1).

[0100] Z1=A+B (1)

[0101] Among them, Z1 represents the vector group, A represents the first word vector, and B represents the second word vector.

[0102] The second method is to combine the first word vector and the third word vector to obtain a vector combination;

[0103] The first word vector and the third word vector can be combined by the embedded expression shown in formula (2).

[0104] Z1=A+C (2)

[0105] Among them, Z1 represents the vector group, A represents the first word vector, and C represents the third word vector.

[0106] The third method is to combine the first word vector, the second word vector and the third word vector to obtain a vector combination;

[0107] The first word vector, the second word vector and the third word vector can be combined by the embedded expression shown in formula (3).

[0108] Z1=A+B+C (3)

[0109] Among them, Z1 represents the vector group, A represents the first word vector, B represents the second word vector, and C represents the third word vector.

[0110] Here, by combining the vectors, searching for the first information associated with the target word from the vector set has the following advantages:

[0111] First, the context information of the proper nouns in the audio data content is converted into corresponding word vectors, and then the word vectors are combined. The combined vector group is used to search in the vector set, and at least two entity vectors that are most related to the proper noun can be directly found. Compared with the related art method of first using text search to find multiple entities matching the literal meaning of the text from the knowledge graph database, and then disambiguating according to the context, this method is simpler to implement and has a faster retrieval speed.

[0112] In one example, taking the simultaneous interpretation scenario as an example, a schematic diagram of the implementation process of searching for the first information associated with the target word from the vector set is described, such as Figure 5 As shown, including:

[0113] Step 1: Combine the first word vector of the target word with the second word vector of the first word to obtain a vector group;

[0114] Assume that the recognized text corresponding to the audio data is "Among many mobile phone brands, I like Apple the most." The target word is "Apple", and the first word is the word above "Apple", namely "mobile phone".

[0115] Combine the first word vector of "apple" with the second word vector of "mobile phone" to get a vector group.

[0116] Step 2: searching for the first vector matching the vector group from the vector set;

[0117] The first vector is composed of two entity vectors that satisfy the embedded expression; the two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of a second knowledge node associated with the first knowledge node. The first vector can be a row vector or a column vector;

[0118] The attribute information of the first knowledge node may include an identifier (ID) of the first knowledge node, such as ID1; the attribute information of the second knowledge node may include an identifier (ID) of the second knowledge node, such as ID2.

[0119] Step 3: Based on the two entity vectors corresponding to the first vector, two knowledge nodes corresponding to the two entity vectors are searched from a knowledge graph database; the knowledge graph database stores the relationship between knowledge nodes and entity vectors;

[0120] The identifiers of the two knowledge nodes corresponding to the two entity vectors are searched in the knowledge graph database, and are ID1 and ID2 respectively.

[0121] Step 4: Acquire the knowledge information of the two knowledge nodes; use the acquired knowledge information of the two knowledge nodes as the first information associated with the target word.

[0122] The knowledge information of knowledge node ID1 is "a communication tool"; the knowledge information of knowledge node ID2 is "a phone produced by an American company". After information processing, the first information associated with "Apple" is "a communication tool produced by an American company". Figure 6 shown.

[0123] In this example, by utilizing the proper nouns in the content of the audio data and the words in the context of the proper nouns, the interpretation information of the proper nouns in the content of the audio data is retrieved from the vector set, so that the unique interpretation information of the proper nouns can be directly determined, which not only eliminates the need to perform the disambiguation step, but also helps users accurately understand the content of the audio data.

[0124] In one example, taking the simultaneous interpretation scenario as an example, a schematic diagram of the implementation process of searching for the first information associated with the target word from the vector set is described, such as Figure 7 As shown, including:

[0125] Step 1: Combine the first word vector of the target word with the third word vector of the second word to obtain a vector group;

[0126] Assume that the recognized text corresponding to the audio data is "Among many mobile phone brands, I like Apple the most." The target word is "Apple", and the second word is the word following "Apple", namely "brand".

[0127] Combine the first word vector of "apple" with the third word vector of "brand" to get a vector group.

[0128] Step 2: searching for the first vector matching the vector group from the vector set;

[0129] The first vector is composed of two entity vectors that satisfy the embedded expression; the two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of a second knowledge node associated with the first knowledge node. The first vector can be a row vector or a column vector;

[0130] The attribute information of the first knowledge node may include an identifier (ID) of the first knowledge node, such as ID1; the attribute information of the second knowledge node may include an identifier (ID) of the second knowledge node, such as ID3.

[0131] Step 3: Based on the two entity vectors corresponding to the first vector, two knowledge nodes corresponding to the two entity vectors are searched from a knowledge graph database; the knowledge graph database stores the relationship between knowledge nodes and entity vectors;

[0132] The identifiers of the two knowledge nodes corresponding to the two entity vectors are searched in the knowledge graph database, and are ID1 and ID3 respectively.

[0133] Step 4: Acquire the knowledge information of the two knowledge nodes; use the acquired knowledge information of the two knowledge nodes as the first information associated with the target word.

[0134] The knowledge information of knowledge node ID1 is "a communication tool"; the knowledge information of knowledge node ID3 is "consumers' awareness of the product". Through information processing, the first information associated with "apple" is "a communication tool with a high degree of consumer awareness", such as Figure 8 shown.

[0135] In this example, by utilizing the proper nouns in the content of the audio data and the words following the proper nouns, the interpretation information of the proper nouns in the content of the audio data is retrieved from the vector set, so that the unique interpretation information of the proper nouns can be directly determined, which not only eliminates the need to perform the disambiguation step, but also helps the user to accurately understand the content of the audio data.

[0136] In one example, taking the simultaneous interpretation scenario as an example, a schematic diagram of the implementation process of searching for the first information associated with the target word from the vector set is described, such as Fig. 9 As shown, including:

[0137] Step 1: Combine the first word vector of the target word, the second word vector of the first word, and the third word vector of the second word to obtain a vector group;

[0138] Assume that the recognized text corresponding to the audio data is "Among many mobile phone brands, I like Apple the most". The target word is "Apple", the first word is the word above "Apple", i.e. "mobile phone". The second word is the word below "Apple", i.e. "brand".

[0139] Combine the first word vector of "apple", the second word vector of "mobile phone", and the third word vector of "brand" to get a vector group.

[0140] Step 2: searching for the first vector matching the vector group from the vector set;

[0141] The first vector is composed of three entity vectors that satisfy the embedded expression; the three entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of two second knowledge nodes that have an association relationship with the first knowledge node. The first vector can be a row vector or a column vector;

[0142] The attribute information of the first knowledge node may include an identifier (ID) of the first knowledge node, such as ID1; the attribute information of the two second knowledge nodes may include identifiers (IDs) of the second knowledge nodes, such as ID2 and ID3.

[0143] Step 3: Based on the three entity vectors corresponding to the first vector, three knowledge nodes corresponding to the three entity vectors are searched from a knowledge graph database; the knowledge graph database stores the relationship between knowledge nodes and entity vectors;

[0144] The identifiers of the two knowledge nodes corresponding to the three entity vectors are searched in the knowledge graph database, namely ID1, ID2 and ID3.

[0145] Step 4: Acquire the knowledge information of the three knowledge nodes; use the acquired knowledge information of the three knowledge nodes as the first information associated with the target word.

[0146] The knowledge information of knowledge node ID1 is "a communication tool"; the knowledge information of knowledge node ID2 is "a phone produced by an American company"; the knowledge information of knowledge node ID3 is "the degree of consumer awareness of the product". Through information processing, the first information associated with "Apple" is "a mobile phone produced in the United States with a high degree of consumer awareness", such as Fig.10 shown.

[0147] In this example, by utilizing the proper nouns in the content of the audio data and the words in the context of the proper nouns, the interpretation information of the proper nouns in the content of the audio data is retrieved from the vector set, so that the unique interpretation information of the proper nouns can be directly determined, which not only eliminates the need to perform the disambiguation step, but also helps the user to accurately understand the content of the audio data.

[0148] In actual application, if users are asked to read information with a lot of text, it will increase the user's understanding burden. In this way, after retrieving the explanation information of the proper nouns contained in the content of the audio data from the vector set, the retrieved information can also be filtered and processed to make the information more concise.

[0149] Based on this, in one embodiment, the method further includes:

[0150] Based on a preset filtering rule, exclude the second information that meets the preset condition in the first information;

[0151] A simultaneous interpretation result is generated by using the third information in the first information except the second information and the recognized text.

[0152] Among them, the preset filtering rule can be that the number of words contained in the additional information is greater than a preset word count threshold, for example, the number of words contained in the additional information is counted; when the counted number of words exceeds 100 words, the first information with a lower importance level in the additional information is deleted to ensure that the number of words in the additional information is within 100 words.

[0153] In actual application, the first information may be translated into information corresponding to the target language of the generator of the audio data, and then a simultaneous interpretation result may be generated based on the translated first information and the translated text.

[0154] Based on this, in one embodiment, the step of generating a simultaneous interpretation result by using the first information and the recognized text includes:

[0155] Receive an acquisition request; the acquisition request is used to acquire simultaneous interpretation data; the acquisition request at least includes: a target language;

[0156] Using a translation model, the first information is translated into fourth information corresponding to the target language; and the recognition text is translated into a translation text corresponding to the target language;

[0157] A simultaneous interpretation result is generated based on the fourth information and the translated text.

[0158] Specifically, a request for obtaining sent by a client may be received.

[0159] In actual application, a simultaneous interpretation result can also be generated based on the first information corresponding to the source language of the speaker of the audio data, the fourth information corresponding to the target language of the generator of the audio data, and the translated text.

[0160] Based on this, in one embodiment, the step of generating a simultaneous interpretation result by using the first information and the recognized text includes:

[0161] Receive an acquisition request; the acquisition request is used to acquire simultaneous interpretation data; the acquisition request at least includes: a target language;

[0162] Using a translation model, the first information is translated into fourth information corresponding to the target language; and the recognition text is translated into a translation text corresponding to the target language;

[0163] A simultaneous interpretation result is generated based on the first information, the fourth information and the translated text.

[0164] Specifically, a request for obtaining sent by a client may be received.

[0165] In actual application, the simultaneous interpretation result can be output in the form of an audio stream using the first information and the translated text.

[0166] Based on this, in one implementation, the outputting of simultaneous interpretation results includes:

[0167] Performing speech synthesis on the simultaneous interpretation result obtained by using the first information and the translation text to synthesize simultaneous interpretation data;

[0168] The synthesized simultaneous interpretation data is sent to the client; the simultaneous interpretation data is used for playing by the client.

[0169] In actual application, the simultaneous interpretation result can be output in the form of an audio stream using the fourth information obtained by simplifying the first information and the translated text.

[0170] Based on this, in one implementation, the outputting of simultaneous interpretation results includes:

[0171] Performing speech synthesis on the simultaneous interpretation result obtained by using the fourth information and the translation text to synthesize simultaneous interpretation data;

[0172] The synthesized simultaneous interpretation data is sent to the client; the simultaneous interpretation data is used for playing by the client.

[0173] Here, the simultaneous interpretation data may be played through the earphone of the client, thereby helping the user using the first terminal to understand the conference content of the speaker of the audio data.

[0174] In actual application, a structured format such as a table, a graph, a web page, etc. can also be used to generate a simultaneous interpretation result based on the first information and the translation text.

[0175] Based on this, in one implementation, the outputting of simultaneous interpretation results includes:

[0176] The simultaneous interpretation result obtained by using the first information and the translated text is sent to a display screen associated with the client; the simultaneous interpretation result is used by the client to display the first information in a first display mode on the display screen, and to display the translated text in a second display mode on the display screen.

[0177] Here, the first display mode may be to display the knowledge information of at least two knowledge nodes corresponding to the first information in the form of a chart, etc. The second display mode may be to display the translated text in the form of a display box.

[0178] In actual application, a structured format such as a table, a graph, a web page, etc. can also be used to generate a simultaneous interpretation result based on the fourth information obtained by simplifying the first information and the translation text.

[0179] Based on this, in one implementation, the outputting of simultaneous interpretation results includes:

[0180] The simultaneous interpretation result obtained by using the fourth information and the translated text is sent to a display screen associated with the client; the simultaneous interpretation result is used by the client to display the fourth information in a first display mode on the display screen, and to display the translated text in a second display mode on the display screen.

[0181] Fig.11 This is a schematic diagram of outputting simultaneous interpretation results, such as Fig.11 As shown, the client can display the knowledge information of at least two knowledge nodes corresponding to the first information in the simultaneous interpretation result in the form of a chart at the upper position of the associated display screen, and the upper position is centered directly above, centered to the right, centered to the left, etc.; the client can display the translated text in the simultaneous interpretation result in a display box at the lower position of the display screen, and the lower position is centered directly below, centered to the right, centered to the left, etc.; the display box can include at least one of the following: a text box, a rich text box.

[0182] By adopting the technical solution of the embodiment of the present application, in order to avoid the need to perform a disambiguation step in the related art when searching for the interpretation information of the target word from the knowledge graph database, the embodiment of the present application adopts the following steps to obtain audio data; recognize the audio data to obtain a recognized text; extract the target word in the recognized text; and extract the first word and the second word of the target word in the recognized text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word; determine the first word vector of the target word; and determine at least one of the second word vector of the first word and the third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, searching the first information associated with the target word from the vector set; the first information is used to interpret the meaning of the target word; using the first information and the recognition text, generating a simultaneous interpretation result; outputting the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data; wherein the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node that has an association relationship with the first knowledge node. By adopting the technical solution of the embodiment of the present application, the first information associated with the target word can be searched from the vector set based on the first word vector, and at least one of the second word vector and the third word vector. Compared with the method of searching the interpretation information of the target word from the knowledge graph database in the related art, there is no need to perform the disambiguation step, the translation accuracy is high, and the implementation method is simple. In this way, it can help users fully and accurately understand the content of the speaker's speech and reduce the difficulty of users to understand the content of the speech.

[0183] In order to implement the method of the embodiment of the present application, the embodiment of the present application also provides a data processing device; Fig.12 As shown, the data processing device includes:

[0184] An acquisition unit 121, configured to acquire audio data;

[0185] The first processing unit 122 is used to recognize the audio data to obtain a recognized text; extract a target word from the recognized text; and extract a first word and a second word of the target word from the recognized text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word;

[0186] The second processing unit 123 is used to determine a first word vector of the target word; and determine at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, search for first information associated with the target word from a vector set; the first information is used to interpret the meaning of the target word;

[0187] The third processing unit 124 is used to generate a simultaneous interpretation result using the first information and the recognized text; and output the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data;

[0188] Among them, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors that satisfy the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node that has an associated relationship with the first knowledge node.

[0189] In actual application, the display device may specifically be a client.

[0190] In one embodiment, the first processing unit 122 is specifically used to: extract at least two keywords from the recognized text; for each of the at least two keywords, perform entity recognition on the corresponding keyword to obtain at least two entity words; search for proper nouns matching the at least two entity words from a dictionary database; and use the entity words corresponding to the found proper nouns as the target words.

[0191] In one embodiment, the second processing unit 123 is specifically used to: combine the first word vector with the fourth word vector to obtain at least one vector group; the fourth word vector is at least one of the second word vector and the third word vector; search for a first vector matching the at least one vector group from a vector set; based on at least two entity vectors corresponding to the first vector, search for at least two knowledge nodes corresponding to the at least two entity vectors from a knowledge graph database; the knowledge graph database stores the relationship between knowledge nodes and entity vectors; obtain knowledge information of the at least two knowledge nodes; and use the obtained knowledge information of the at least two knowledge nodes as the first information associated with the target word.

[0192] In one embodiment, when generating simultaneous interpretation results, the third processing unit 124 is specifically used to: exclude the second information that meets the preset conditions in the first information based on the preset filtering rules; and generate the simultaneous interpretation results using the third information in the first information except the second information and the recognized text.

[0193] In one embodiment, the third processing unit 124 is specifically used to: receive an acquisition request; the acquisition request is used to obtain simultaneous interpretation data; the acquisition request at least includes: a target language; using a translation model to translate the first information into fourth information corresponding to the target language; and translating the recognized text into a translation text corresponding to the target language; and generating a simultaneous interpretation result based on the fourth information and the translation text.

[0194] Among them, the acquisition request sent by the client can be received.

[0195] In one embodiment, the third processing unit 124 is specifically used to: perform speech synthesis on the simultaneous interpretation result obtained by using the fourth information and the translation text to synthesize simultaneous interpretation data; send the synthesized simultaneous interpretation data to the client; and the simultaneous interpretation data is used for playback by the client.

[0196] In one embodiment, the third processing unit 124 is specifically used to: send the simultaneous interpretation result obtained using the fourth information and the translated text to a display screen associated with the client; the simultaneous interpretation result is used by the client to display the fourth information in a first display mode on the display screen, and to display the translated text in a second display mode on the display screen.

[0197] In actual application, the acquisition unit 121 can be implemented through a communication interface; the first processing unit 122, the second processing unit 123, and the third processing unit 124 can all be implemented by a processor in the device.

[0198] It should be noted that: when the device provided in the above embodiment performs data processing, only the division of the above program modules is used as an example. In actual applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the terminal is divided into different program modules to complete all or part of the processing described above. In addition, the device provided in the above embodiment and the data processing method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0199] Based on the hardware implementation of the above device, the embodiment of the present application also provides an electronic device, Fig.13 Schematic diagram of the hardware structure of the electronic device of the embodiment of the present application. Fig.13 As shown, the electronic device 130 includes a memory 133, a processor 132, and a computer program stored in the memory 133 and executable on the processor 132; when the processor 132 executes the program, the method provided by one or more of the above technical solutions is implemented.

[0200] It should be noted that the specific steps implemented when the processor 132 executes the program have been described in detail above and will not be repeated here.

[0201] It is understood that the electronic device 130 also includes a communication interface 131, and the communication interface 131 is used to exchange information with other devices; at the same time, the various components in the electronic device 130 are coupled together through a bus system 134. It is understood that the bus system 134 is configured to achieve connection and communication between these components. In addition to the data bus, the bus system 134 also includes a power bus, a control bus, and a status signal bus.

[0202] It can be understood that the memory 133 in this embodiment can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disk, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), and direct RAM bus random access memory (DRRAM, Direct Rambus Random Access Memory).The memories described in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0203] The method disclosed in the above embodiment of the present application can be applied to the processor 132, or implemented by the processor 132. The processor 132 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 132 or the instruction in the form of software. The above processor 132 may be a general-purpose processor, a DSP, or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 132 can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. In combination with the steps of the method disclosed in the embodiment of the present application, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium, which is located in a memory, and the processor 132 reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0204] The present application also provides a storage medium, specifically a computer storage medium, more specifically a computer readable storage medium, on which computer instructions, namely a computer program, are stored, and when the computer instructions are executed by a processor, the method provided by one or more of the above technical solutions is implemented.

[0205] In the several embodiments provided by the present invention, it should be understood that the disclosed methods and intelligent devices can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0206] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0207] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0208] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium, which, when executed, executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, disks or optical disks.

[0209] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0210] It should be noted that: "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0211] In addition, the technical solutions described in the embodiments of the present application can be combined arbitrarily without conflict.

[0212] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A data processing method, comprising: Get audio data; Recognize the audio data to obtain recognized text; Extracting target words from the recognized text; and extracting a first word and a second word of the target word in the recognition text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word; the target word is an entity word corresponding to a proper noun; The proper noun is a word that matches an entity word corresponding to a keyword in the recognized text that is found in a dictionary database; Determine a first word vector of the target word; and determine at least one of a second word vector of the first word and a third word vector of the second word; based on the first word vector, and at least one of the second word vector and the third word vector, search for first information associated with the target word from a vector set; The first information is used to explain the meaning of the target word; Using the first information and the recognized text, generating a simultaneous interpretation result; Outputting the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data; Wherein, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors satisfying the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node associated with the first knowledge node; The searching, based on the first word vector and at least one of the second word vector and the third word vector, for first information associated with the target word from a vector set includes: Combining the first word vector with the fourth word vector to obtain at least one vector group; the fourth word vector is at least one of the second word vector and the third word vector; Finding a first vector matching the at least one vector group from the vector set; Based on the at least two entity vectors corresponding to the first vector, searching for at least two knowledge nodes corresponding to the at least two entity vectors from a knowledge graph database; the knowledge graph database stores the relationship between the knowledge nodes and the entity vectors; Acquire knowledge information of the at least two knowledge nodes; and use the acquired knowledge information of the at least two knowledge nodes as first information associated with the target word.

2. The method according to claim 1, wherein: The extracting the target words in the recognized text includes: extracting at least two keywords from the recognized text; For each of the at least two keywords, perform entity recognition on the corresponding keyword to obtain at least two entity terms; Searching for proper nouns matching the at least two entity words from a dictionary database; The entity words corresponding to the found proper nouns are used as the target words.

3. The method according to claim 1, wherein: When generating the simultaneous interpretation result, the method further includes: Based on a preset filtering rule, exclude the second information that meets the preset condition in the first information; A simultaneous interpretation result is generated by using the third information in the first information except the second information and the recognized text.

4. The method according to any one of claims 1 to 3, wherein: The step of generating a simultaneous interpretation result by using the first information and the recognized text includes: Receive an acquisition request; the acquisition request is used to acquire simultaneous interpretation data; the acquisition request at least includes: a target language; Using a translation model, the first information is translated into fourth information corresponding to the target language; and the recognition text is translated into a translation text corresponding to the target language; A simultaneous interpretation result is generated based on the fourth information and the translated text.

5. The method according to claim 4, wherein: The output simultaneous interpretation result includes: Performing speech synthesis on the simultaneous interpretation result obtained by using the fourth information and the translation text to synthesize simultaneous interpretation data; The synthesized simultaneous interpretation data is sent to the client; the simultaneous interpretation data is used for playing by the client.

6. The method according to claim 4, wherein: The output simultaneous interpretation result includes: The simultaneous interpretation result obtained by using the fourth information and the translated text is sent to a display screen associated with the client; the simultaneous interpretation result is used by the client to display the fourth information in a first display mode on the display screen, and to display the translated text in a second display mode on the display screen.

7. A data processing device, comprising: An acquisition unit, used for acquiring audio data; A first processing unit, used for recognizing the audio data to obtain a recognized text; Extracting target words from the recognized text; and extracting a first word and a second word of the target word in the recognition text; the first word is a word in the previous context of the target word; the second word is a word in the following context of the target word; the target word is an entity word corresponding to a proper noun; The proper noun is a word that matches an entity word corresponding to a keyword in the recognized text that is found in a dictionary database; A second processing unit is used to determine a first word vector of the target word; and determine a second word vector of the first word and a third word vector of the second word; Based on the first word vector and at least one of the second word vector and the third word vector, searching for first information associated with the target word from a vector set; The first information is used to explain the meaning of the target word; A third processing unit, configured to generate a simultaneous interpretation result using the first information and the recognized text; Outputting the simultaneous interpretation result; the simultaneous interpretation result is used to be presented on a display device when playing the audio data; Wherein, the vector set includes at least one first vector; each first vector is composed of at least two entity vectors satisfying the embedded expression; the at least two entity vectors represent the attribute information of the first knowledge node in the knowledge graph database and the attribute information of at least one second knowledge node associated with the first knowledge node; The second processing unit is used to combine the first word vector with the fourth word vector to obtain at least one vector group; the fourth word vector is at least one of the second word vector and the third word vector; search for a first vector matching the at least one vector group from a vector set; based on at least two entity vectors corresponding to the first vector, search for at least two knowledge nodes corresponding to the at least two entity vectors from a knowledge graph database; the knowledge graph database stores the relationship between knowledge nodes and entity vectors; Acquire knowledge information of the at least two knowledge nodes; and use the acquired knowledge information of the at least two knowledge nodes as first information associated with the target word.

8. An electronic device, comprising: a processor and a memory for storing a computer program capable of being executed on the processor; Wherein, when the processor is used to run the computer program, it executes the steps of the method described in any one of claims 1 to 6.

9. A storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images