Voice processing method, apparatus, intelligent agent, storage medium and program product

Abstract

A voice processing method, an apparatus, an intelligent agent, a storage medium and a program product are provided, relating to the field of human-computer voice interaction. The method comprises: acquiring a first voice segment, an intent of the first voice segment being a navigation intent, and the first voice segment comprising first address information; acquiring a second voice segment, the time interval between the second voice segment and the first voice segment being greater than or equal to a first threshold and less than or equal to a second threshold; and, on the basis of a semantic relationship between the second voice segment and the first voice segment, determining whether to concatenate first text and second text, the first text being text content corresponding to the first voice segment, and the second text being text content corresponding to the second voice segment. Whether to concatenate the texts corresponding to the two voice segments is determined by means of semantic analysis, such that text more aligned with a user's intent can be obtained, and a result better matching the user's intent can be output, thereby improving user experience.

Description

A speech processing method, device, intelligent agent, storage medium, and program product. Technical Field

[0001] This application relates to the field of human-computer voice interaction, and in particular to a voice processing method, device, intelligent agent, storage medium, and program product. Background Technology

[0002] Voice recognition technology has been widely used in smart home devices and smart vehicle devices, enabling a natural human-computer voice interaction experience.

[0003] Automatic speech recognition (ASR) is a crucial technology in current speech recognition capabilities. ASR requires identifying valid speech segments within an audio signal, involving the detection of beginning and end points (speech start and end points). Current speech end point detection techniques utilize voice activity detection (VAD), which considers speech to have ended after detecting a fixed duration of silence. However, in some scenarios (such as voice navigation), users need to pause and think before speaking completely, causing VAD to prematurely identify the end of speech. This results in incomplete text recognition, leading to outputs that do not meet user expectations and a poor user experience. Summary of the Invention

[0004] This application provides a speech processing method, apparatus, intelligent agent, storage medium, and program product, with the aim of obtaining text that is closer to the user's intent, thereby outputting results that better match the user's intent.

[0005] In a first aspect, this application provides a speech processing method that can be applied to an intelligent agent. The steps of the method can be executed by the intelligent agent, or the method can be executed by a component (such as a chip, chip system, speech processing device, etc.) configured in the intelligent agent, or it can be implemented by a logic module or software capable of realizing all or part of the functions of the intelligent agent. This application does not limit the scope of the method.

[0006] In this application, intelligent agents may include vehicles, smartphones, smartwatches, smart bracelets, robots, or drones, etc.

[0007] For example, the method includes: acquiring a first voice segment, the intent of which is navigation, and the first voice segment includes first address information; acquiring a second voice segment, the time interval between the second voice segment and the first voice segment being greater than or equal to a first threshold and less than or equal to a second threshold, the second threshold being greater than the first threshold; and determining whether to concatenate a first text and a second text based on the semantic relationship between the second voice segment and the first voice segment, the first text being the text content corresponding to the first voice segment and the second text being the text content corresponding to the second voice segment.

[0008] Based on the above technical content, by performing semantic analysis on two adjacent speech segments to determine whether to concatenate the text corresponding to these two speech segments, text that is closer to the user's expectations can be obtained, thereby outputting results that better meet the user's expectations and thus improving the user experience.

[0009] In conjunction with the first aspect, in some possible implementations, determining whether to concatenate the first text and the second text based on the semantic relationship between the second speech segment and the first speech segment includes: concatenating the first text and the second text to obtain a third text when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hyponym of the first address entity; or, not concatenating the first text and the second text when the intent of the second speech segment is not a navigation intent, or the second speech segment does not include second address information; or, not concatenating the first text and the second text when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hypernym of the first address entity. Alternatively, if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is greater than or equal to a third threshold, then the first text and the second text are not concatenated; or, if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold, then the first text and the second text are concatenated to obtain a third text; wherein the first address entity is the last address entity in the first address information, and the second address entity is the first address entity in the second address information.

[0010] In conjunction with the first aspect, in some possible implementations, the method further includes: displaying a first navigation search result derived based on the third text.

[0011] In conjunction with the first aspect, in some possible implementations, the method includes: displaying the third text.

[0012] In conjunction with the first aspect, in some possible implementations, the method further includes: displaying a second navigation search result, which is obtained based on the second text.

[0013] In conjunction with the first aspect, in some possible implementations, after obtaining the first speech segment, the method further includes: processing the first speech segment to obtain the first text; and displaying a third navigation search result, which is obtained based on the first text.

[0014] In conjunction with the first aspect, in some possible implementations, the method further includes: acquiring a third speech segment, wherein the time interval between the third speech segment and the second speech segment is greater than or equal to the first threshold and less than or equal to the second threshold; and determining whether to concatenate a fourth text and the third text based on the semantic relationship between the third speech segment and the second speech segment, wherein the fourth text is the text content corresponding to the third speech segment.

[0015] Secondly, this application provides a voice processing apparatus, including modules or units for implementing the methods of the first aspect and any possible implementation thereof. It should be understood that each module or unit can implement its corresponding function by executing a computer program.

[0016] Thirdly, this application provides a voice processing apparatus, including a processor that can be used to execute the methods in the first aspect and any possible implementation of the first aspect.

[0017] The device may further include a memory for storing instructions and data. The memory is coupled to the processor, which, when executing the instructions stored in the memory, can implement the methods described in the foregoing aspects. The device may also include a communication interface for communicating with other devices; exemplary, the communication interface may be a transceiver, circuit, bus, module, or other type of communication interface.

[0018] Fourthly, this application provides an intelligent agent comprising execution steps for performing the first aspect and any possible implementation thereof. The intelligent agent includes corresponding modules for performing the above-described methods. The modules included in the intelligent agent can be implemented in software and / or hardware.

[0019] Fifthly, this application provides an intelligent agent including a processor. The processor is coupled to a memory and can be used to execute a program in the memory to implement the execution steps in the first aspect and any possible implementation thereof.

[0020] Optionally, the agent may also include a memory.

[0021] Optionally, the agent also includes a communication interface, with the processor coupled to the communication interface.

[0022] Optionally, in this application, the intelligent agent may include a vehicle, smartphone, smartwatch, smart bracelet, robot, or drone, etc.

[0023] In a sixth aspect, this application provides a chip system including at least one processor for supporting the implementation of the functions involved in the first aspect and any possible implementation of the first aspect, such as receiving or processing data and / or information involved in the above methods.

[0024] In one possible design, the chip system also includes a memory for storing program instructions and data, which may be located within or outside the processor.

[0025] The chip system can consist of chips or include chips and other discrete components.

[0026] In a seventh aspect, this application provides a computer-readable storage medium including a computer program that, when run on a computer, causes the computer to implement the methods of the first aspect and any possible implementation of the first aspect.

[0027] Eighthly, this application provides a computer program product comprising: a computer program (also referred to as code or instructions) that, when run, causes a computer to perform the methods of the first aspect and any possible implementation thereof.

[0028] It should be understood that the second to eighth aspects of this application correspond to the technical solutions of the first aspect of this application, and the beneficial effects achieved by each aspect and the corresponding feasible implementation are similar, and will not be repeated here. Attached Figure Description

[0029] Figure 1 is a schematic diagram of speech recognition based on VAD;

[0030] Figure 2 is a schematic structural block diagram of a voice processing device provided in an embodiment of this application;

[0031] Figure 3 is a schematic flowchart of the speech processing method provided in an embodiment of this application;

[0032] Figure 4 is a flowchart illustrating the decision-making process for whether to splice the first text and the second text, as provided in this application.

[0033] Figure 5 is another flowchart illustrating the decision-making process for whether to splice the first text and the second text, as provided in this application.

[0034] Figure 6 is another schematic flowchart of the speech processing method provided in the embodiments of this application;

[0035] Figure 7 is another schematic flowchart of the speech processing method provided in the embodiments of this application;

[0036] Figure 8 is another schematic block diagram of the voice processing device provided in an embodiment of this application;

[0037] Figure 9 is another schematic block diagram of the voice processing device provided in the embodiments of this application. Detailed Implementation

[0038] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0039] First, in this application, the terms “comprising” and “having” and any variations thereof are intended to cover non-exclusive inclusion, for example, an apparatus, system, product or device that includes a series of modules, units or units is not necessarily limited to those modules, units or units that are explicitly listed, but may include other modules, units or units that are not explicitly listed or that are inherent to such apparatus, system, product or device.

[0040] Second, in this application, the words "exemplarily" and "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design that is described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of words such as "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0041] Third, in this application, "when," "under the circumstances," "if," and "if" all refer to the device making a corresponding action under certain objective circumstances, and are not time-limited, nor do they require the device to make a judgment action when it is implemented, nor do they mean that there are other limitations.

[0042] Fourth, in this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, "first threshold," "second threshold," and "third threshold" are used to distinguish different thresholds, and "first text," "second text," "third text," and "fourth text" are used to distinguish different texts, without limiting their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply that they are different.

[0043] Fifth, in this application, "preset" can be understood as predefined, defined, pre-defined, stored, pre-stored, pre-negotiated, or pre-configured, etc.

[0044] Sixth, in this application, "at least one (kind, type)" refers to one (kind, type) or multiple (kinds, types). "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship. The specific meaning can be understood in conjunction with the context.

[0045] Seventh, in this application, information C is used to determine information D, including both determining information D based solely on information C and determining it based on information C and other information. Furthermore, information C can also be used to determine information D indirectly, for example, in the case where information D is determined based on information E, and information E is determined based on information C.

[0046] To facilitate understanding of the embodiments of this application, some technical terms or vocabulary involved in this application will be briefly explained below.

[0047] 1. ASR (Automatic Speech Recognition): Designed to convert spoken language into written text. The core task of an ASR system is to analyze and process speech signals to identify words and phrases. This technology plays an important role in many applications, including voice assistants, voice input methods, automatic caption generation, and voice translation.

[0048] The main components and technologies of an ASR system may include, but are not limited to, the following:

[0049] 1) Speech signal processing: Preprocessing the input speech signal to remove noise and extract features. This process can use Mel-frequency cepstral coefficients (MFCCs) or other acoustic features.

[0050] 2) Acoustic Model: Matching the features of the speech signal with phonemes (the basic units of speech). Acoustic models can be trained using, but are not limited to, deep neural networks (DNNs), convolutional neural networks (CNNs), or recurrent neural networks (RNNs).

[0051] 3) Language Model: Utilizing statistical methods to predict the probability of word sequences, thereby improving recognition accuracy. Language models help ASR systems understand context in order to more accurately identify words.

[0052] 4) Decoder: Combining acoustic and language models, a search algorithm is used to determine the most likely word sequence. This is a crucial step in converting speech signals into text.

[0053] 5) Training data: The ASR system requires a large amount of labeled speech data for training in order to improve its recognition ability and accuracy.

[0054] 2. VAD: VAD is used to detect and distinguish speech activity from non-speech activity (such as silence, background noise, etc.) in audio signals. VAD plays an important role in speech processing systems, especially in saving bandwidth, improving speech recognition accuracy, and enhancing the quality of voice communication.

[0055] The main functions of VAD may include, but are not limited to, the following:

[0056] 1) Save bandwidth and storage: In voice communication systems, VAD can reduce the amount of audio data transmitted and stored because it only transmits or records audio when voice activity is detected.

[0057] 2) Improve speech recognition accuracy: In an ASR system, VAD can help the ASR system focus on processing the actual speech part, ignoring silence and noise, thereby improving the accuracy and efficiency of recognition.

[0058] 3) Speech enhancement and noise reduction: By identifying speech and non-speech segments, VAD can help speech enhancement systems remove background noise more effectively and improve audio quality.

[0059] The technical implementation of VAD can include, but is not limited to, the following:

[0060] 1) Energy detection: Determines the presence of speech activity based on the energy level of the audio signal. Simple but easily affected by background noise.

[0061] 2) Zero Crossing Rate (ZCR): Speech activity is detected by analyzing the zero crossing rate of the audio signal. Speech signals typically have a higher ZCR.

[0062] 3) Spectral analysis: using spectral features (such as spectral entropy, spectral centroid, etc.) to distinguish between speech and non-speech segments.

[0063] 4) Machine learning methods: Using trained models (such as neural networks) to detect more complex speech activities, which can perform well in noisier environments.

[0064] 5) Adaptive algorithm: Dynamically adjusts detection thresholds and parameters to adapt to different environments and background noise levels.

[0065] Through these technologies, the VAD system can effectively detect speech activity in various application scenarios, thereby improving the overall performance of the speech processing system.

[0066] 3. Semantic Understanding: In the field of Natural Language Processing (NLP), semantic understanding refers to the process by which computer systems or algorithms analyze and understand the deeper meaning of natural language text. This involves more than just analyzing the literal meaning of words; it can also include understanding one or more of the following: contextual relationships, implied meanings, semantic relationships, the intent expressed in the text, etc. The goal of semantic understanding is to enable machines to understand and interpret the meaning of language in a way that is similar to that of humans.

[0067] Semantic understanding may involve one or more of the following:

[0068] 1) Lexical semantics: Understanding the meaning of words, including the different meanings of synonyms, antonyms, and polysemous words.

[0069] 2) Grammatical and semantic role labeling: Identify the grammatical structure of a sentence and the roles of different words in the sentence.

[0070] 3) Contextual understanding: Use contextual information to eliminate ambiguity and understand the correct meaning of polysemous words and phrases.

[0071] 4) Entity recognition and relation extraction: Identify entities (such as names of people and places) in the text and the relationships between them.

[0072] 5) Sentiment analysis: Analyzing the emotions or attitudes expressed in the text.

[0073] 6) Intent recognition: Identifying the intent or purpose expressed in the text, especially in dialogue systems.

[0074] In this application, semantic understanding can be replaced by semantic parsing, semantic analysis, natural language parsing, concept extraction, contextual understanding, etc., and this application does not limit it.

[0075] 4. Edit Distance (ED): This refers to the minimum number of operations required to transform one string into another. Operations can include adding (adding a character), deleting (deleting a character), and modifying (changing a character). For example, the minimum edit distance between string A and string B can be denoted as ED. AB Then the edit distance similarity between string A and string B satisfies:

[0076] (For ease of description, this formula will be referred to as Formula 1), where L A L represents the length of string A. B This indicates the length of string B.

[0077] Voice recognition (VAR) technology is widely used in smart home and in-vehicle devices, enabling natural human-computer voice interaction. Automatic Speech Recognition (ASR) is a crucial technique in current VAR technology. ASR requires identifying valid speech segments within an audio signal, involving the detection of the beginning and end points of speech. Among current VAR detection techniques, the Voice End Point Detection (VAD) method is used, which considers the speech to have ended after detecting a silence of a fixed duration.

[0078] Figure 1 is a schematic diagram of speech recognition based on VAD.

[0079] Figure 1 shows a speech segment. Based on VAD, the speech segment can be identified, the start position of the speech segment can be detected, and the speech can be considered to have ended after a silence of a fixed duration (e.g., 600 milliseconds, 800 ms, etc.). As shown in Figure 1, the speech is considered to have timed out 600 ms after the last speech frame of speech segment 1.

[0080] However, in some scenarios (such as voice navigation), users need to pause and think before they can finish speaking. This can cause the VAD method to prematurely assume that the speech has ended, resulting in incomplete text recognition and outputting results that do not match the user's intention, leading to a poor user experience.

[0081] To address the aforementioned technical issues, this application provides a speech processing method that performs semantic analysis on two adjacent speech segments to determine whether to concatenate the text corresponding to these two speech segments. This method can produce text that is closer to the user's expectations, thereby outputting results that better meet the user's expectations and improving the user experience.

[0082] Before detailing the speech processing method provided in the embodiments of this application, the speech processing device applied to the embodiments of this application will be described exemplarily with reference to FIG2.

[0083] Figure 2 is a schematic structural block diagram of a voice processing device provided in an embodiment of this application.

[0084] As shown in Figure 2, the speech processing device applicable to this application may include a data acquisition module, a data processing module, and a decision module. The data acquisition module may include, but is not limited to, a microphone, and can be used to acquire audio data streams (e.g., speech data). The data processing module can be used to acquire acoustic information from the data acquisition module and convert the acoustic information into text information. The decision module can be used to perform semantic understanding on the text information from the data processing module and make a splicing decision based on the result of the semantic understanding.

[0085] In one possible implementation, the voice processing device may further include an execution module that can be used to perform corresponding actions based on the text or instructions output by the decision module.

[0086] The speech processing method provided in the embodiments of this application will be described exemplarily below with reference to the accompanying drawings.

[0087] Figure 3 is a schematic flowchart of the speech processing method provided in the embodiments of this application.

[0088] The method 300 shown in Figure 3 may include steps 310 to 330.

[0089] The steps of this method can be executed by an intelligent agent, or by a component (such as a chip, chip system, etc.) configured in the intelligent agent, or by a logic module or software capable of implementing all or part of the functions of the intelligent agent, or by a speech processing device. This application does not limit the scope of the method. The following describes the steps of method 300 in detail using the example of a speech processing device executing the method 300.

[0090] In step 310, a first voice segment is obtained, the intent of which is navigation intent, and the first voice segment includes first address information.

[0091] For example, the speech processing device can acquire speech segments based on the data acquisition module shown in FIG2, for instance, it can acquire a first speech segment. After acquiring the first speech segment, the speech processing device can convert the first speech segment into first text based on the data processing module shown in FIG2. Further, the speech processing device can perform semantic understanding on the first text based on the decision module shown in FIG2.

[0092] As an example and not a limitation, in a voice navigation scenario, the voice processing device performs semantic understanding on the first text, analyzes whether the intent of the first text is a navigation intent, and can analyze whether the first text includes address information (for ease of description, this address information is referred to as the first address information).

[0093] In this application, the intent of the first text is taken as a navigation intent, and the first text includes the first address information as an example for description. That is, in this application, the intent of the first voice segment is a navigation intent, and the first voice segment includes the first address information.

[0094] For example, the first address information is BJ City, CY District.

[0095] In step 320, a second speech segment is obtained, wherein the time interval between the second speech segment and the first speech segment is greater than or equal to a first threshold and less than or equal to a second threshold.

[0096] It is understood that the first threshold is a preset threshold in the VAD method, based on which the end position of a speech segment can be identified. That is, based on the first threshold, an audio segment can be divided into a first speech segment and a second speech segment. As an example and not a limitation, in practical applications, the first threshold can be 500ms, 600ms, 700ms, or 800ms, etc., and this application does not impose any limitation on the specific value of the first threshold.

[0097] The first threshold can also be preset. Based on this second threshold, it can be ensured that the second speech segment and the first speech segment are in the same round of human-computer dialogue. In practical application scenarios, the second threshold is greater than the first threshold. The second threshold can be 1 minute, 1.5 minutes, or 2 minutes, etc. This application does not limit the specific value of the second threshold.

[0098] For example, the speech processing device can acquire speech segments based on the data acquisition module shown in FIG2, for instance, it can acquire a second speech segment. After acquiring the second speech segment, the speech processing device can convert the second speech segment into second text based on the data processing module shown in FIG2. Further, the speech processing device can perform semantic understanding on the second text based on the decision module shown in FIG2.

[0099] As an example and not a limitation, in a voice navigation scenario, the voice processing device performs semantic understanding on the second text, analyzing whether the intent of the second text is a navigation intent, and whether the second text includes address information (for ease of description, this address information is referred to as the second address information). That is, the voice processing device can analyze whether the intent of the second voice segment is a navigation intent, and whether the second voice segment includes the second address information.

[0100] For example, the second address information is RM Road FY Community.

[0101] In step 330, based on the semantic relationship between the second speech segment and the first speech segment, it is determined whether to concatenate the first text and the second text. The first text is the text content corresponding to the first speech segment, and the second text is the text content corresponding to the second speech segment.

[0102] For example, the speech processing device can perform semantic understanding on the first speech segment and the second speech segment, and make a splicing decision based on the semantic relationship between the first speech segment and the second speech segment to determine whether to splice the first text and the second text.

[0103] In one possible implementation, determining whether to concatenate the first text and the second text based on the semantic relationship between the second speech segment and the first speech segment includes: concatenating the first text and the second text to obtain a third text if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a hyponym of the first address entity; or, not concatenating the first text and the second text if the intent of the second speech segment is not navigation, or the second speech segment does not include second address information; or, not concatenating the first text and the second text if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a hypernym of the first address entity; or... If the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is greater than or equal to a third threshold, then the first text and the second text are not concatenated; or, if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold, then the first text and the second text are concatenated to obtain a third text; wherein the first address entity is the last address entity in the first address information, and the second address entity is the first address entity in the second address information.

[0104] For example, in this application, the intent of a speech segment (or text) can be determined based on currently known intent recognition methods. As an example and not a limitation, for instance, a speech processing device can, based on semantic understanding (a submodule), recognize that the speech segment (or text) contains expressions such as "navigation," "navigate to," "navigate to," or "go to *** (address information)," or, if address information is present, recognize that the intent of the speech segment is navigation.

[0105] For example, if the first address information is HD District of BJ City, then HD District is the first address entity; if the second address information is WD Plaza of HD District, then HD District is the second address entity.

[0106] In this application, address entities can be divided into 14 levels (or classes) of elements, as shown in Table 1 below. It is understood that Table 1 is only an example, and in actual application scenarios, address entities can be divided into more or fewer levels (or classes). This application does not limit the number of levels to which address entities can be divided.

[0107] Table 1

[0108] As shown in Table 1, the lower level (or class number or ranking) is the hyponym of the higher level, and correspondingly, the higher level is the supernym of the lower level. Two address entities with the same level are appositives.

[0109] For example, ZH country, HB province, SJZ city, where ZH country (level 1) is a superordinate term of HB province (level 2), and HB province is a hyponym of ZH country; HB province is a superordinate term of SJZ city, and SJZ city is a hyponym of HB province.

[0110] For example, in SJZ City CHA District, SJZ City (level 3) is a superordinate term of CHA District (level 4), and CHA District is a hyponym of SJZ City; in SJZ City GY County, SJZ City is a superordinate term of GY County (level 4), and GY County is a hyponym of SJZ City; CHA District and GY County are appositives.

[0111] For example, in QD City, CHY District, QD City (level 3) is a superordinate term of CHY District (level 4), and CHY District is a subordinate term of QD City; in CHY District, GSHY Market, CHY District is a superordinate term of GSHY Market (a point of interest, level 12), and GSHY Market is a subordinate term of CHY District; it can be understood that QD City is also a superordinate term of GSHY Market, and GSHY Market is a subordinate term of QD City.

[0112] For example, in the phrase "YD Hotel LDM Store", "YD Hotel" (a point of interest, level 12) is a superordinate term of "LDM Store" (a branch, level 13), and "LDM Store" is a subordinate term of "YD Hotel".

[0113] For example, in "LC Building in NJ City," NJ City is a superordinate term of LC Building (a commercial office building, a point of interest, level 12), and LC Building is a hyponym of NJ City. Similarly, in "LZH Building in JIN City," JIN City is a superordinate term of LZH Building (a commercial office building, a point of interest, level 12), and LZH Building is a hyponym of JIN City. NJ City and JIN City are appositives, as are LC Building and LZH Building. It can be understood that NJ City is also a superordinate term of LZH Building, and LZH Building is a hyponym of NJ City; JIN City is also a superordinate term of LC Building, and LC Building is a hyponym of JIN City.

[0114] For example, in "JN District JN Traditional Chinese Medicine Hospital", "JN District" is a superordinate term of "JN Traditional Chinese Medicine Hospital" (a point of interest, level 12), and "JN Traditional Chinese Medicine Hospital" is a hyponym of "JN District"; in "JN District DQ Garden", "JN District" is a superordinate term of "DQ Garden" (a residential community, a point of interest, level 12), and "DQ Garden" is a hyponym of "JN District"; "JN Traditional Chinese Medicine Hospital" and "DQ Garden" are appositives.

[0115] For example, in SHB Culture Center Area F, SHB Culture Center (a point of interest, level 12) is a superordinate term of Area F (a sub-point of interest, level 14), and Area F is a hyponym of SHB Culture Center; in HQ Airport Terminal 2, HQ Airport (a point of interest, level 12) is a superordinate term of Terminal 2 (a sub-point of interest, level 14), and Terminal 2 is a hyponym of HQ Airport; in GQ Forest Park East Gate 2, GQ Forest Park (a point of interest, level 12) is a superordinate term of East Gate 2 (a sub-point of interest, level 14), and East Gate 2 is a hyponym of GQ Forest Park.

[0116] Figure 4 is a flowchart illustrating the decision-making process for whether to splice the first text and the second text, as provided in this application.

[0117] In step 401, the second voice segment is a navigation intent and includes second address information.

[0118] For example, the voice processing device can analyze the intent of the second voice segment, determine whether the second voice segment is a navigation intent, and determine whether the second voice segment includes address information. It is understood that, in order to distinguish it from the address information included in the first voice segment, the address information included in the second voice segment is recorded as the second address information.

[0119] If the second voice segment is a navigation intention and includes second address information, the voice processing device may execute step 402; if the intention of the second voice segment is not a navigation intention, or if the second voice segment does not include second address information, the voice processing device may execute step 406.

[0120] In step 402, is the second address entity a hyponym of the first address entity?

[0121] For example, when the second voice segment is a navigation intention and the second voice segment includes second address information, the voice processing device can determine whether the first address entity in the second address information (referred to as the second address entity for ease of description) is a hyponym of the last address entity in the first address information (referred to as the first address entity for ease of description).

[0122] If the second address entity is a hyponym of the first address entity, the speech processing device may execute step 405; if the second address entity is not a hyponym of the first address entity, the speech processing device may execute step 403.

[0123] In step 403, is the second address entity an appositive of the first address entity?

[0124] For example, if the second address entity is not a hyponym of the first address entity, the speech processing device can determine whether the second address entity is a synonym of the first address entity.

[0125] If the second address entity is a synonym of the first address entity, the speech processing device may execute step 404; if the second address entity is not a synonym of the first address entity (that is, if the second address entity is a superordinate of the first address entity), the speech processing device may execute step 406.

[0126] In step 404, is the edit distance similarity between the first address entity and the second address entity less than the third threshold?

[0127] For example, when the second address entity is an appositive of the first address entity, the speech processing device can calculate the edit distance between the first address entity and the second address entity. Furthermore, when the second address entity is an appositive of the first address entity, the speech processing device can calculate the edit distance similarity between the first address entity and the second address entity based on the edit distance between the first address entity and the second address entity.

[0128] In this application, edit distance can refer to the edit distance of the pinyin of the address entity (hereinafter referred to as pinyin edit distance) or the edit distance of the Chinese text of the address entity (hereinafter referred to as Chinese edit distance). That is, in practical application scenarios, one example is that the pinyin edit distance between the first address entity and the second address entity can be calculated, and the similarity of the pinyin edit distance between the first address entity and the second address entity can be calculated based on the pinyin edit distance; another example is that the Chinese edit distance between the first address entity and the second address entity can be calculated, and the similarity of the Chinese edit distance between the first address entity and the second address entity can be calculated based on the Chinese edit distance. This application does not limit this.

[0129] It is understood that the edit distance and edit distance similarity between the first address entity and the second address entity can be calculated based on methods known in current NLP for calculating edit distance and edit distance similarity, and this application does not limit this.

[0130] For example, the pinyin of the first address entity can be denoted as string A, and the pinyin of the second address entity can be denoted as string B. Then, the pinyin edit distance between the first address entity and the second address entity can be denoted as ED. AB Furthermore, the similarity of the pinyin edit distance between the first address entity and the second address entity can be calculated according to Formula 1 mentioned in the technical terminology section above.

[0131] Without loss of generality, the process of calculating the similarity of the Chinese edit distance between the first address entity and the second address entity can be the same as the process of calculating the similarity of the Pinyin edit distance between the first address entity and the second address entity. For the sake of brevity, it will not be elaborated here.

[0132] After calculating the similarity of the pinyin edit distance between the first address entity and the second address entity, the speech processing device can determine whether the edit distance similarity between the first address entity and the second address entity is less than a third threshold. For example, the third threshold can be 40%, 45%, 50%, 55%, or 60%, etc., and the third threshold can be preset. This application does not limit the specific value of the third threshold.

[0133] If the edit distance similarity between the first address entity and the second address entity is less than the third threshold, the speech processing device may execute step 405; if the edit distance similarity between the first address entity and the second address entity is not less than the third threshold (that is, if the edit distance similarity between the first address entity and the second address entity is greater than or equal to the third threshold), the speech processing device may execute step 406.

[0134] Step 405: Concatenate the first text and the second text to obtain the third text.

[0135] For example, if the second speech segment is a navigation intent and includes second address information speech, and the second address entity is a hyponym of the first address entity, the processing device can concatenate the first text and the second text to obtain a third text. Alternatively, if the intent of the second speech segment is a navigation intent, the second speech segment includes second address information, the second address entity is a hyponym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold, the processing device can concatenate the first text and the second text to obtain a third text.

[0136] In step 406, the first text and the second text are not concatenated.

[0137] It is understandable that step 406 can also be interpreted as not concatenating the first and second texts, but instead directly performing navigation search based on the second text.

[0138] Figure 5 is another flowchart illustrating the decision-making process for whether to splice the first text and the second text, as provided in this application.

[0139] As shown in Figure 5, the difference from Figure 4 lies in the execution order of steps 402 and 403. More specifically, when the second voice segment is a navigation intention and includes second address information, the voice processing device can first execute step 403, that is, first determine whether the second address entity is a synonym of the first address entity. If the second address entity is a synonym of the first address entity, the voice processing device can execute step 404; if the second address entity is not a synonym of the first address entity, the voice processing device can execute step 402, that is, determine whether the second address entity is a hyponym of the first address entity. If the second address entity is a hyponym of the first address entity, the voice processing device can execute step 405; if the second address entity is not a hyponym of the first address entity (that is, if the second address entity is a superordinate of the first address entity), the voice processing device can execute step 406.

[0140] Optionally, the method 300 further includes: displaying the third text.

[0141] For example, after concatenating the first text and the second text to obtain the third text, the voice processing device can display the third text on a user interface (UI). For instance, the third text can be displayed on a navigation search interface. Or, for example, the third text can be displayed within a navigation search box. This provides the user with a better interactive experience.

[0142] It is understandable that in practical application scenarios, the third text may not be displayed on the UI interface, and this application does not impose any restrictions on this.

[0143] In one possible implementation, the method 300 further includes: displaying a first navigation search result derived based on the third text.

[0144] For example, after concatenating the first text and the second text to obtain the third text, the voice processing device can perform a navigation search based on the third text to obtain a navigation search result (for ease of description, this navigation search result is referred to as the first navigation search result). After obtaining the first navigation search result, the voice processing device can display the first navigation search result on the UI.

[0145] In one possible implementation, the method 300 further includes: displaying a second navigation search result, which is obtained based on the second text.

[0146] For example, after the voice processing device decides not to concatenate the first text and the second text, it can directly perform a navigation search based on the second text to obtain a navigation search result (for ease of description, this navigation search result is referred to as the second navigation search result). After obtaining the second navigation search result, the voice processing device can display the second navigation search result on the UI.

[0147] Optionally, after obtaining the first speech segment, the method 300 further includes: processing the first speech segment to obtain the first text; and displaying a third navigation search result, which is obtained based on the first text.

[0148] For example, after acquiring a first speech segment, the voice processing device can process the first speech segment to obtain a first text. Further, the voice processing device can perform a navigation search based on the first text to obtain a navigation search result (for ease of description, this navigation search result is referred to as the third navigation search result). After obtaining the third navigation search result, the voice processing device can display the third navigation search result on the UI.

[0149] In one possible scenario, when the voice processing device acquires the second voice segment, it may not have had time to display the third navigation search result.

[0150] In one possible implementation A, if the second voice segment is obtained earlier than the third navigation search result is displayed, the third navigation search result can be displayed first, followed by the second navigation search result or the first navigation search result.

[0151] In one possible implementation B, if the second voice segment is obtained earlier than the third navigation search result is displayed, the third navigation search result may not be displayed; instead, the second or first navigation search result may be displayed directly.

[0152] Figure 6 is another schematic flowchart of the speech processing method provided in the embodiments of this application.

[0153] In step 601, the semantic understanding (submodule) obtains the first speech segment: Navigate to GD City.

[0154] For example, after the user says the first voice segment "Navigate to GD City", the data acquisition module of the voice processing device can acquire the first voice segment. The data acquisition module can then pass the first voice segment to the semantic understanding sub-module in the data processing module. Accordingly, the semantic understanding sub-module acquires the first voice segment.

[0155] In step 602, the semantic understanding (submodule) performs semantic understanding on the first speech segment.

[0156] For example, the semantic understanding (submodule) can perform semantic understanding on the first speech segment to obtain the semantic understanding result. For example, it can analyze that the text information (i.e. the first text) corresponding to the first speech segment is "navigate to GD City", the intent of the first speech segment is navigation, and the slot (i.e. the address information) of the first speech segment is GD City.

[0157] In step 603, the semantic understanding (submodule) passes the semantic understanding result to the concatenation decision (submodule). Accordingly, the concatenation decision (submodule) obtains the semantic understanding result from the semantic understanding (submodule).

[0158] In step 604, the splicing decision (submodule) caches the intent and slot for this round and makes a splicing judgment.

[0159] It's understandable that the splicing decision (submodule) can cache the intent and slot for the current round to facilitate splicing judgments in the current round and subsequent rounds. It's also understandable that the intent and slot for the current round refer to the intent and slot of the first speech segment.

[0160] In step 605, the splicing decision (submodule) feeds back the splicing decision result to the semantic understanding (submodule), and the splicing decision result is no splicing.

[0161] In this round of dialogue, there are no other audio segments before the first audio segment, so they are not spliced ​​together.

[0162] In step 606, the semantic understanding (submodule) passes "Slot: GD City" to the map (module). Accordingly, the map (module) obtains "Slot: GD City".

[0163] It is understandable that "Slot: GD City" here means that the first address information is GD City. After obtaining "Slot: GD City", the map (module) can perform navigation search based on "GD City", that is, it can perform navigation search based on the first address information.

[0164] In step 607, the map (module) feeds back navigation search results to the semantic understanding (submodule).

[0165] Understandably, the navigation search results in this feedback are the same as the third navigation search results mentioned earlier.

[0166] In step 608, the semantic understanding (submodule) asks the user which number to choose.

[0167] Understandably, the semantic understanding (submodule) can present navigation search results to the user and ask the user which navigation search result to select.

[0168] It is also understood that, in another possible implementation, steps 607 and 608 can be combined into one step, whereby the map (module) presents the navigation search results to the user and asks the user which navigation search result to select. This application does not limit this.

[0169] In step 609, the semantic understanding (submodule) obtains the second speech segment: Navigate to JZSZ Ecological Farm.

[0170] For example, after the user says the second voice segment "JZSZ Ecological Farm", the data acquisition module of the voice processing device can acquire the second voice segment. The data acquisition module can then pass the second voice segment to the semantic understanding sub-module in the data processing module. Accordingly, the semantic understanding sub-module acquires the second voice segment.

[0171] In step 610, the semantic understanding (submodule) performs semantic understanding on the second speech segment.

[0172] For example, the semantic understanding (submodule) can perform semantic understanding on the second speech segment to obtain the semantic understanding result. For example, it can analyze that the text information (i.e., the second text) corresponding to the second speech segment is "JZSZ Ecological Farm", the intent of the second speech segment is navigation, and the slot (i.e., the address information) of the first speech segment is JZSZ Ecological Farm.

[0173] In step 611, the semantic understanding (submodule) passes the semantic understanding result to the splicing decision (submodule). Accordingly, the splicing decision (submodule) obtains the semantic understanding result from the semantic understanding (submodule).

[0174] In step 612, the splicing decision (submodule) caches the intent and slot for this round and makes a splicing judgment.

[0175] It is understandable that the splicing decision (submodule) can cache the intent and slot of the current round to facilitate splicing judgment in the current round and in subsequent rounds. It is also understandable that the intent and slot of the current round refer to the intent and slot of the second speech segment.

[0176] In step 613, the splicing decision (submodule) feeds back the splicing decision result to the semantic understanding (submodule), and the splicing decision result is splicing.

[0177] For example, if the semantic relationship between the second speech segment and the first speech segment satisfies the splicing condition, the splicing decision (submodule) can provide a splicing decision result for splicing the first text and the second text.

[0178] It is understood that the splicing decision conditions may include the aforementioned "the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hyponym of the first address entity" and "the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold".

[0179] After splicing the first and second texts, we get "GD City JZSZ Ecological Farm". In this example, the third text is "GD City JZSZ Ecological Farm", and the spliced ​​slot is "GD City JZSZ Ecological Farm".

[0180] In step 614, the semantic understanding (submodule) passes "Slot: GD City JZSZ Ecological Farm" to the map (module). Accordingly, the map (module) obtains "Slot: GD City JZSZ Ecological Farm City".

[0181] After obtaining the location "Slot: GD City JZSZ Ecological Farm", the map (module) can perform a navigation search based on "GD City JZSZ Ecological Farm" and obtain the first navigation search result.

[0182] In step 615, the map (module) feeds back navigation search results to the semantic understanding (submodule).

[0183] It is understandable that the navigation search results in this feedback are the same as the first navigation search results mentioned above.

[0184] In step 616, the semantic understanding (submodule) asks the user which number to choose.

[0185] Understandably, the semantic understanding (submodule) can present navigation search results to the user and ask the user which navigation search result to select.

[0186] It is also understood that, in another possible implementation, steps 615 and 616 can be combined into one step, whereby the map (module) presents the navigation search results to the user and asks the user which navigation search result to select. This application does not limit this.

[0187] Figure 7 is another schematic flowchart of the speech processing method provided in the embodiments of this application.

[0188] The difference between Figure 6 and the speech processing method shown in Figure 7 is that Figure 6 includes steps 607 and 608, which means displaying the third navigation search results, while Figure 7 does not include steps 607 and 608, which means the third navigation search results are not displayed.

[0189] Steps 601 to 606 and steps 609 to 615 in Figure 7 are the same as those in Figure 6. For a detailed description of the specific steps, please refer to the relevant description of Figure 6 above. For the sake of brevity, they will not be repeated here.

[0190] Using the speech processing method provided in this application, not only can the effective concatenation of addresses be achieved, but also erroneous concatenation can be avoided. More detailed examples are provided below in conjunction with Tables 2 and 3.

[0191] Table 2

[0192] As shown in Table 2, if the first and second texts are not concatenated, the navigation search results will not meet the user's expectations.

[0193] More specifically, as in example number 1, if the user is not currently in the CHY district, without concatenating the first and second texts, the search will return several GSHY markets closer to the user's current location, and may not return any GSHY markets in the CHY district. In other words, without concatenation, the navigation search results do not meet the user's expectations. However, after using the speech processing method provided in this application to make a concatenation decision, the first and second texts are concatenated, and the resulting address information is "GSHY market in the CHY district". The navigation search based on this address information will better meet or be closer to the user's expectations.

[0194] As in example number 2, without concatenation, the search would return several YD Hotel branches near the user's current location. The Laodongmen branch might not appear in the first few search results, resulting in navigation search results that do not meet the user's expectations. However, after using the speech processing method provided in this application to make a concatenation decision, the first and second texts are concatenated, and the resulting address information is "YD Hotel Laodongmen Branch". The navigation search based on this address information will better meet or be closer to the user's expectations.

[0195] Table 3

[0196] As shown in Table 3, Examples 3 and 4 are examples where no concatenation is performed after the speech processing method of this application makes a concatenation decision. Based on the hierarchical relationship of address elements as shown in Table 1, the first and second texts are concatenated when the first address entity in the second text is a hyponym of the last address entity in the first text; the first and second texts are not concatenated when the first address entity in the second text is a hypernym of the last address entity in the first text; and the first address entity in the second text is an appositive of the last address entity in the first text, and the edit distance similarity is used to determine whether the first and second texts are concatenated, thus avoiding incorrect concatenation results.

[0197] In one possible implementation, the method 300 further includes: acquiring a third speech segment, wherein the time interval between the third speech segment and the second speech segment is greater than or equal to the first threshold and less than or equal to the second threshold; and determining whether to concatenate a fourth text and the third text based on the semantic relationship between the third speech segment and the second speech segment, wherein the fourth text is the text content corresponding to the third speech segment.

[0198] It is understandable that in real-world applications, users may need to express the address information of the place they want to go in multiple rounds of conversation.

[0199] For example, in a single dialogue, if the speech processing device acquires the following speech segments: first, "Navigate to SJZ City"; second, "WD Plaza, CHA District"; and third, "East Gate"; then, based on the semantic relationship between the second and first speech segments, the speech processing device can decide whether to concatenate the first and second texts to obtain the third text, "Navigate to WD Plaza, CHA District, SJZ City". Furthermore, based on the semantic relationship between the third and second speech segments, the speech processing device can determine whether to concatenate the fourth and third texts.

[0200] It is understandable that the detailed process of determining whether to concatenate the fourth and third texts based on the semantic relationship between the third and second speech segments can be found in the description above regarding determining whether to concatenate the first and second texts based on the semantic relationship between the second and first speech segments. For brevity, this will not be repeated here. For example, the voice processing device determines that the fourth and third texts need to be concatenated based on the semantic relationship between the third and second speech segments, thus obtaining the fifth text "Navigate to the East Gate of WD Plaza, CHA District, SJZ City". Furthermore, the voice processing device can perform a navigation search based on the address information "East Gate of WD Plaza, CHA District, SJZ City", thereby obtaining navigation search results that better meet the user's expectations.

[0201] For example, in a dialogue, the voice processing device acquires the following voice segments: first, "Navigate to YY Road"; second, "Lane 123"; third, "Building 42"; and fourth, "Room 401". Based on the semantic relationship between the second and first voice segments, the voice processing device can decide to concatenate the first and second texts to obtain the third text, "Navigate to Lane 123, YY Road". Further, based on the semantic relationship between the third and second voice segments, the voice processing device can determine to concatenate the fourth text corresponding to the third voice segment with the third text to obtain the fifth text, "Navigate to Building 42, Lane 123, YY Road". Further still, based on the semantic relationship between the fourth and third voice segments, the voice processing device can determine to concatenate the sixth text corresponding to the fourth voice segment with the fifth text to obtain the seventh text, "Navigate to Room 401, Building 42, Lane 123, YY Road".

[0202] Based on the above technical solution, by considering the semantic relationship between two adjacent speech segments, not only can the address be effectively concatenated, but also incorrect concatenation can be avoided.

[0203] The speech processing method provided in this application has been described in detail above with reference to the accompanying drawings. The speech processing apparatus provided in the embodiments of this application will be described in detail below with reference to Figures 8 and 9.

[0204] It should be understood that the voice processing device shown in Figures 8 and 9 can be used to implement the functions of the voice processing device in the above method embodiments, and therefore can also achieve the beneficial effects of the above method embodiments. In the embodiments of this application, the voice processing device can be an intelligent agent, or a component (such as a chip, chip system, processor, etc.) configured in the intelligent agent, or a logic module or software capable of implementing some or all of the functions of the intelligent agent.

[0205] Figure 8 is another schematic block diagram of the voice processing device provided in the embodiments of this application.

[0206] As shown in Figure 8, the voice processing device 800 includes an acquisition module 810 and a processing module 820. The voice processing device 800 can be used to implement the functions of the voice processing device in any of the method embodiments shown in Figures 3 to 7.

[0207] For example, when the voice processing device 800 is used to implement the function of the voice processing device in the method embodiment shown in FIG3, the acquisition module 810 can be used to acquire a first voice segment, the intent of which is a navigation intent, and the first voice segment includes first address information; the acquisition module 810 can also be used to acquire a second voice segment, the time interval between which is greater than or equal to a first threshold and less than or equal to a second threshold, the second threshold being greater than the first threshold; the processing module 820 can be used to determine whether to concatenate a first text and a second text based on the semantic relationship between the second voice segment and the first voice segment, the first text being the text content corresponding to the first voice segment, and the second text being the text content corresponding to the second voice segment.

[0208] Optionally, the processing module 820 can be specifically used to: concatenate the first text and the second text to obtain a third text when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hyponym of the first address entity; or, when the intent of the second speech segment is not a navigation intent, or the second speech segment does not include second address information, not concatenate the first text and the second text; or, when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hypernym of the first address entity, not concatenate the first text and the second text; or, when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hypernym of the first address entity, not concatenate the first text and the second text; or, when the intent of the second speech segment is a navigation intent, and the second speech segment includes second address information, and the second address entity is a hypernym of the first address entity, not concatenate the first text and the second text. The second speech segment includes second address information, and the second address entity is a synonym of the first address entity. If the edit distance similarity between the first address entity and the second address entity is greater than or equal to a third threshold, the first text and the second text are not concatenated. Alternatively, if the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a synonym of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold, the first text and the second text are concatenated to obtain a third text. Wherein, the first address entity is the last address entity in the first address information, and the second address entity is the first address entity in the second address information.

[0209] Optionally, the processing module 820 can also be used to: display a first navigation search result, which is derived based on the third text.

[0210] Optionally, the processing module 820 can also be used to display the third text.

[0211] Optionally, the processing module 820 can also be used to: display a second navigation search result, which is obtained based on the second text.

[0212] Optionally, the processing module 820 can also be used to: process the first speech segment to obtain the first text; and display a third navigation search result, which is obtained based on the first text.

[0213] Optionally, the processing module 820 can also be used to: acquire a third speech segment, wherein the time interval between the third speech segment and the second speech segment is greater than or equal to the first threshold and less than or equal to the second threshold; and determine whether to concatenate a fourth text and the third text based on the semantic relationship between the third speech segment and the second speech segment, wherein the fourth text is the text content corresponding to the third speech segment.

[0214] For a more detailed description of each of the above modules, please refer directly to the relevant description in the method embodiment shown in Figure 3, which will not be repeated here.

[0215] It should be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0216] Figure 9 is another schematic block diagram of the voice processing device provided in the embodiments of this application.

[0217] The voice processing device 900 can be a chip system, or a device configured with a chip system to implement the methods described in the above-described method embodiments. In this application embodiment, the chip system can be composed of chips, or it can include chips and other discrete devices.

[0218] As shown in FIG9, the voice processing device 900 may include a processor 910, which can be used to execute computer programs or instructions in memory to implement the steps performed by the voice processing device in any of the method embodiments shown in FIG3 to FIG7.

[0219] Optionally, the voice processing device 900 further includes a communication interface 920. The communication interface 920 can be used to communicate with other devices via a transmission medium, thereby enabling the voice processing device 900 to communicate with other devices. The communication interface 920 can be, for example, a transceiver, interface, bus, circuit, or a device capable of transmitting and receiving functions. The processor 910 can use the communication interface 920 to input and output data, and to implement the method described in any of the embodiments corresponding to Figures 3 to 7. Specifically, the voice processing device 900 can be used to implement the functions of the voice processing device in the above-described method embodiments.

[0220] Optionally, the voice processing device 900 further includes at least one memory 930 for storing program instructions and / or data. The memory 930 is coupled to the processor 910. The coupling in this embodiment is an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, used for information exchange between devices, units, or modules. The processor 910 may operate in conjunction with the memory 930. The processor 910 may execute program instructions stored in the memory 930.

[0221] In this application, the memory 930 can be integrated into the processor 910, or the processor 910 and the memory 930 can be set up separately; this application does not limit this.

[0222] It should be understood that the coupling in the embodiments of this application is an indirect coupling or communication connection between devices, units, or modules, which can be electrical, mechanical, or other forms, used for information interaction between devices, units, or modules. The processor 910 may operate in conjunction with the memory 930. The specific connection medium between the processor 910, communication interface 920, and memory 930 is not limited in the embodiments of this application. In Figure 9, the processor 910, communication interface 920, and memory 930 are connected via a bus 940. The bus 940 is represented by a thick line in Figure 9. The connection methods between other components are only illustrative and not intended to be limiting. The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used in Figure 9, but this does not indicate that there is only one bus or one type of bus.

[0223] This application also provides an intelligent agent, which includes the voice processing device as described above.

[0224] This application also provides an intelligent agent, which includes modules that can implement the methods described in any of the embodiments shown in FIG3 to FIG7.

[0225] Optionally, in the embodiments of this application, the intelligent agent may include a vehicle, smartphone, smartwatch, smart bracelet, robot, or drone, etc.

[0226] This application also provides a chip system including at least one processor for supporting the implementation of the functions involved in any possible implementation of the embodiments shown in Figures 3 to 7, such as receiving or processing data and / or instruction information involved in the above methods.

[0227] In one possible design, the chip system also includes a memory for storing program instructions and data, which may be located within or outside the processor.

[0228] The chip system can consist of chips or include chips and other discrete components.

[0229] This application also provides a computer program product, which includes a computer program (also referred to as code or instructions) that, when run, can implement the steps performed by the intelligent agent in any of the embodiments shown in Figures 3 to 7.

[0230] This application also provides a computer-readable storage medium storing a computer program (also referred to as code or instructions). When the computer program is run, it can implement the steps performed by the intelligent agent in any of the embodiments shown in Figures 3 to 7.

[0231] It should be understood that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method embodiments can be completed by the integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0232] It should also be understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0233] The terms "unit," "module," etc., used in this specification can be used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. In the embodiments of this application, "unit" and "module" have the same meaning and can be used interchangeably.

[0234] Those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. In the several embodiments provided in this application, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0235] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0236] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0237] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital video discs, DVDs), or semiconductor media (e.g., solid-state disks, SSDs), etc.

[0238] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the technology, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0239] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A speech processing method, characterized in that, The method includes: Obtain a first voice segment, the intent of which is a navigation intent, and the first voice segment includes first address information; Obtain a second speech segment, wherein the time interval between the second speech segment and the first speech segment is greater than or equal to a first threshold and less than or equal to a second threshold, wherein the second threshold is greater than the first threshold; Based on the semantic relationship between the second speech segment and the first speech segment, it is determined whether to concatenate the first text and the second text, where the first text is the text content corresponding to the first speech segment and the second text is the text content corresponding to the second speech segment.

2. The method according to claim 1, characterized in that, The step of determining whether to concatenate the first text and the second text based on the semantic relationship between the second speech segment and the first speech segment includes: If the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is a hyponym of the first address entity, then the first text and the second text are concatenated to obtain the third text; or, If the intent of the second voice segment is not navigation, or if the second voice segment does not include the second address information, the first text and the second text will not be concatenated; or, If the intent of the second voice segment is navigation, and the second voice segment includes second address information, and the second address entity is a hypernym of the first address entity, then the first text and the second text are not concatenated; or, If the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is an appositive of the first address entity, and the edit distance similarity between the first address entity and the second address entity is greater than or equal to a third threshold, then the first text and the second text are not concatenated; or, If the intent of the second speech segment is navigation, and the second speech segment includes second address information, and the second address entity is an appositive of the first address entity, and the edit distance similarity between the first address entity and the second address entity is less than the third threshold, then the first text and the second text are concatenated to obtain the third text. Wherein, the first address entity is the last address entity in the first address information, and the second address entity is the first address entity in the second address information.

3. The method according to claim 2, characterized in that, The method further includes: The first navigation search result is displayed, which is derived based on the third text.

4. The method according to claim 2 or 3, characterized in that, The method includes: The third text is displayed.

5. The method according to claim 2, characterized in that, The method further includes: Display a second navigation search result, which is derived from the second text.

6. The method according to any one of claims 1 to 5, characterized in that, After acquiring the first speech segment, the method further includes: The first speech segment is processed to obtain the first text; The third navigation search result is displayed, which is derived from the first text.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: Obtain a third speech segment, wherein the time interval between the third speech segment and the second speech segment is greater than or equal to the first threshold and less than or equal to the second threshold; Based on the semantic relationship between the third speech segment and the second speech segment, it is determined whether to concatenate the fourth text and the third text, wherein the fourth text is the text content corresponding to the third speech segment.

8. A voice processing device, characterized in that, The apparatus includes a module for performing the method as described in any one of claims 1 to 7.

9. A voice processing device, characterized in that, Including processor and memory, among which, The memory is used to store programs; The processor is used to invoke the program so that the device performs the method as described in any one of claims 1 to 7.

10. An intelligent agent, characterized in that, The intelligent agent is a module used to perform the method as described in any one of claims 1 to 7.

11. An intelligent agent, characterized in that, The intelligent agent includes the speech processing device as described in claim 8 or 9.

12. The intelligent agent as described in claim 10 or 11, characterized in that, The intelligent agents include vehicles, smartphones, smartwatches, smart bracelets, robots, or drones.

13. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed, the method as described in any one of claims 1 to 7 is performed.

14. A computer program product, characterized in that, Includes a computer program that, when run, causes the method as described in any one of claims 1 to 7 to be performed.

Images