WO2022007723A1 - 语句类型识别方法、装置、电子设备和存储介质 - Google Patents

语句类型识别方法、装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2022007723A1
WO2022007723A1 PCT/CN2021/104353 CN2021104353W WO2022007723A1 WO 2022007723 A1 WO2022007723 A1 WO 2022007723A1 CN 2021104353 W CN2021104353 W CN 2021104353W WO 2022007723 A1 WO2022007723 A1 WO 2022007723A1
Authority
WO
WIPO (PCT)
Prior art keywords
sentence
semantic representation
representation information
probability
sentence type
Prior art date
Application number
PCT/CN2021/104353
Other languages
English (en)
French (fr)
Inventor
陈佳豪
傅玮萍
丁文彪
刘子韬
Original Assignee
北京世纪好未来教育科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京世纪好未来教育科技有限公司 filed Critical 北京世纪好未来教育科技有限公司
Publication of WO2022007723A1 publication Critical patent/WO2022007723A1/zh
Priority to US18/147,080 priority Critical patent/US11775769B2/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G06F40/216—Parsing using statistical methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/284—Lexical analysis, e.g. tokenisation or collocates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Definitions

  • the present application relates to the field of natural language processing, and in particular, to a sentence type identification method, apparatus, electronic device and storage medium.
  • Teachers will use a variety of teaching methods in the classroom to stimulate students' thinking and study. These teaching methods include asking questions, assigning group work, etc. Among them, questioning is one of the key teaching methods for teachers in classroom teaching. It can increase the interaction between teachers and students, improve the participation of students, enable teachers and students to have open discussions, understand each other's thoughts, and lead to deeper thinking of students.
  • the embodiments of the present application provide a sentence type identification method, device, electronic device and storage medium to solve the problems existing in the related art.
  • the technical solutions are as follows:
  • an embodiment of the present application provides a method for a statement type, including:
  • the sentence type corresponding to the semantic representation information is determined as the sentence type of the target sentence.
  • the plurality of candidate sentence types include non-question sentences and a plurality of question sentence types
  • the sentence type corresponding to the semantic representation information is determined.
  • the sentence type corresponding to the semantic representation information is determined according to the probability corresponding to each candidate sentence type, including:
  • the sentence type corresponding to the maximum probability is determined as the sentence type corresponding to the semantic representation information.
  • the plurality of question types include open-ended questions, intellectual-seeking questions, dialog management questions, and procedural questions.
  • the plurality of candidate sentence types include questions and non-questions
  • the sentence type corresponding to the semantic representation information is a question sentence or a non-question sentence.
  • the semantic representation information includes multiple dimensional information
  • determining the probability that the sentence type corresponding to the semantic representation information is a question sentence includes:
  • the probability that the sentence type corresponding to the semantic representation information is a question sentence is obtained.
  • the method further includes:
  • the corresponding semantic representation information including:
  • Corresponding semantic representation information is obtained according to at least one word information in the target sentence.
  • an embodiment of the present application provides a sentence type identification device, including:
  • the semantic recognition model is used to obtain the corresponding semantic representation information according to the target sentence determined from the classroom teaching speech to be recognized;
  • the classifier is used to select the sentence type corresponding to the semantic representation information from the plurality of candidate sentence types as the sentence type of the target sentence.
  • the plurality of candidate sentence types include non-question sentences and a plurality of question sentence types;
  • the classifier includes a plurality of first probability calculation models and first decision units corresponding to the plurality of candidate sentence types respectively;
  • each first probability calculation model in the plurality of first probability calculation models is used for determining the probability that the sentence type corresponding to the semantic representation information is the candidate sentence type for its corresponding candidate sentence type;
  • the first decision unit is configured to determine the sentence type corresponding to the semantic representation information according to the probability corresponding to each candidate sentence type in the plurality of candidate sentences.
  • the first decision unit is configured to select a maximum probability from the probabilities determined respectively by the plurality of first probability calculation models, and determine the sentence type corresponding to the maximum probability as the sentence type corresponding to the semantic representation information.
  • the plurality of question types include open-ended questions, intellectual-seeking questions, dialog management questions, and procedural questions.
  • the plurality of candidate sentence types include questions and non-questions;
  • the classifier includes a second probability calculation model and a second decision unit;
  • the second probability calculation model is used to determine the probability that the sentence type corresponding to the semantic representation information is a question sentence
  • the second decision unit is configured to determine, according to the probability, whether the sentence type corresponding to the semantic representation information is a question sentence or a non-question sentence.
  • the semantic representation information includes multiple dimension information
  • the second probability calculation model is used to obtain the probability that the sentence type corresponding to the semantic representation information is a question sentence by calculating the inner product of multiple dimension information and normalizing the inner product.
  • the apparatus further includes:
  • a tokenizer which is used to segment the target sentence to obtain at least one word information in the target sentence
  • the semantic recognition model obtains corresponding semantic representation information according to at least one word information in the target sentence.
  • an embodiment of the present application provides an electronic device, where the electronic device includes: a memory and a processor.
  • the memory and the processor communicate with each other through an internal connection path, the memory is used for storing instructions, the processor is used for executing the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor makes the processor The method in any one of the embodiments of the above aspects is performed.
  • embodiments of the present application provide a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program runs on a computer, the method in any one of the implementation manners of the above aspects is executed.
  • the embodiment of the present application first determines the semantic feature information for the target sentence determined from the classroom teaching speech, and then selects the corresponding sentence type based on the semantic representation information, the embodiment of the present application classifies the target sentence based on semantic understanding. It can improve the classification accuracy of target sentences, help to identify diverse sentence types, and help improve the quality of teaching.
  • FIG. 1 is a schematic diagram of a method for identifying a sentence type according to an embodiment of the present application.
  • FIG. 2 is a schematic diagram of a process of processing speech in classroom teaching according to an embodiment of the present application.
  • FIG. 3 is a schematic diagram of a method for identifying a sentence type according to an embodiment of the present application.
  • FIG. 4 is a schematic diagram of a method for identifying a sentence type according to an embodiment of the present application.
  • FIG. 5 is a schematic diagram of a sentence type identification device according to an embodiment of the present application.
  • FIG. 6 is a schematic diagram of a sentence type identification device according to an embodiment of the present application.
  • FIG. 7 is a schematic diagram of a sentence type identification device according to an embodiment of the present application.
  • FIG. 8 is a schematic diagram of a sentence type identification device according to an embodiment of the present application.
  • FIG. 9 is a structural block diagram of an electronic device according to an embodiment of the present application.
  • FIG. 1 shows a schematic diagram of a method for identifying a sentence type according to an embodiment of the present application. As shown in Figure 1, the method may include:
  • Step S101 determining the target sentence from the classroom teaching speech to be recognized
  • Step S102 obtaining corresponding semantic representation information according to the target sentence
  • Step S103 selecting a sentence type corresponding to the semantic representation information from a plurality of candidate sentence types
  • Step S104 determining the sentence type corresponding to the semantic representation information as the sentence type of the target sentence.
  • all sentences or some sentences in classroom teaching speech can be determined as target sentences.
  • the number of target statements can be one or more.
  • the above steps S102 to S104 are respectively executed to obtain a sentence type corresponding to each target sentence.
  • the target statement may be a text-formatted statement.
  • Text recognition can be performed on classroom teaching speech in audio format to obtain target sentences in text format.
  • a VAD Voice Activity Detection, voice activity detection
  • ASR Voice Activity Detection, voice activity detection
  • the target sentence is input into the sentence type identification device that can perform the above steps S102 to S103, and the sentence type of the target sentence, such as type 1, type 2 or type 3, can be obtained.
  • the sentence type recognition model may include a semantic recognition model and a classifier, wherein the semantic recognition model may output corresponding semantic representation information according to the input target sentence or feature information of the target sentence.
  • the multiple candidate sentence types are used as the classification results of the classifier, the semantic representation information is input into the classifier, and the classifier can be used to output one of the multiple candidate sentence types as the sentence type corresponding to the semantic representation vector.
  • the semantic recognition model may include a Transformer model, a BiLSTM (Bi-directional Long Short-Term Memory) model, or a BiGRU (bidirection gated recurrent unit), and the like.
  • the classifier may include one or more MLPs (Multilayer Perceptrons) or CNNs (Convolutional Neural Networks).
  • the method for the statement type may further include:
  • step S102 acquiring corresponding semantic representation information according to the target sentence includes: acquiring corresponding semantic representation information according to at least one word information in the target sentence.
  • the word vector is, for example, a one-hot (one-hot) vector or a distributed word vector.
  • the target sentence is first mapped to a word vector, and the target sentence is represented by machine language, which is beneficial to accurately obtain the feature information of the target sentence.
  • the semantic representation information may also include multiple dimension information. That is, semantic representation information can also be represented by multi-dimensional vectors.
  • each N-dimensional word vector E [CLS] of the target sentence and the corresponding dimension information E 1 to E N are input into a semantic recognition model such as a Transformer (Trm) model.
  • the Transformer model processes each word vector and outputs a semantic representation vector H [CLS] .
  • the semantic representation vector is input to the classifier, and the classifier can output the sentence type corresponding to the semantic representation vector as the sentence type of the target sentence.
  • the embodiment of the present application first determines the semantic feature information for the target sentence determined from the classroom teaching speech, and then selects the corresponding sentence type based on the semantic representation information, the embodiment of the present application classifies the target sentence based on semantic understanding. It can improve the classification accuracy of the target sentence and is conducive to identifying diverse sentence types.
  • questions can be divided into the following types:
  • Discourse-management Questions For example, questions that are not related to the teaching content and are mainly used to attract attention, including “really?”, “right?", “right?", “is it right?” ?”Wait. For example, “Yes, that's the same question I mentioned earlier?", “Yes, right?" or “Because of what? Because... ah”, etc. are not mandatory questions.
  • Procedural Questions For example, questions that are unrelated to the teaching content but related to the teaching procedure, such as questions about reading topics, debugging equipment, greeting students, and asking about recent living conditions. For example: “How are you doing?”, “Can you hear me?”, “Do you still have questions? Let's finish the questions first, and then I'll tell you this", “Did I tell you?", “Tell me Where?” and so on.
  • the multiple candidate sentence types in the above step S103 include non-question sentences and multiple question sentence types.
  • the sentence type corresponding to the semantic representation information selected from multiple candidate sentence types may include:
  • the sentence type corresponding to the semantic representation information is determined.
  • the semantic representation vector H [CLS] is input into the classifier, and the classifier includes multiple probability calculation models such as the MLP in the figure.
  • a DNN model can also be used, such as a DNN model including three convolutional layers of size 128, 64, and 32, respectively.
  • Each probability calculation model corresponds to one candidate sentence type, including a non-question sentence Y 1 and a plurality of question sentence types Y 2 to Y M .
  • Each MLP calculates the probability that the sentence type corresponding to the semantic representation vector H [CLS] is the Y corresponding to the MLP, that is, the probability corresponding to each candidate sentence type Y 1 to Y M can be used to determine the sentence type corresponding to the semantic representation information.
  • the probabilities corresponding to the multiple candidate sentence types may be normalized to convert the probabilities of the multiple candidate sentence types into the same distribution space, so as to improve the determination of the sentence type. accuracy. For example, use softmax to normalize multiple probabilities.
  • determining the sentence type corresponding to the semantic representation information according to the probability corresponding to each candidate sentence type which may include:
  • the sentence type corresponding to the maximum probability is determined as the sentence type corresponding to the semantic representation information.
  • the candidate sentences may also include question sentences and non-question sentences.
  • the sentence type corresponding to the semantic representation information is selected from a plurality of candidate sentence types, including:
  • the sentence type corresponding to the semantic representation information is a question sentence or a non-question sentence.
  • the sentence type corresponding to the semantic representation information is a question sentence
  • the probability is 0.6
  • the probability is greater than the threshold of 0.5
  • the sentence type corresponding to the semantic representation information is determined to be a question sentence.
  • the threshold value is increased or decreased.
  • the probability that the sentence type corresponding to the semantic representation information is a question sentence can be obtained by calculating the inner product of multiple dimensional information and normalizing the inner product.
  • the embodiment of the present application determines the sentence type based on the semantic representation information, the recognition accuracy of the sentence type is improved to a certain extent. Therefore, in the case where only question sentences and non-question sentences need to be identified, the information of multiple dimensions is integrated by calculating the inner product. Summarizing and then normalizing can obtain recognition results with required accuracy and reduce computational complexity.
  • the embodiment of the present application since the embodiment of the present application first determines the semantic feature information for the target sentence determined from the classroom teaching speech, and then selects the corresponding sentence type based on the semantic representation information, the embodiment of the present application is based on the semantic understanding of the target sentence. Classification can improve the classification accuracy of target sentences, help to identify diverse sentence types, and help improve the quality of teaching.
  • FIG. 5 shows a schematic diagram of a sentence type identification device according to an embodiment of the present application.
  • the apparatus may be an electronic device for executing the above sentence type identification method, or may be a sentence type identification model, such as the models shown in FIG. 3 and FIG. 4 .
  • the apparatus may include:
  • the semantic recognition model 201 is used for acquiring corresponding semantic representation information according to the target sentence determined from the speech to be recognized in classroom teaching;
  • the classifier 202 is configured to select the sentence type corresponding to the semantic representation information from the plurality of candidate sentence types as the sentence type of the target sentence.
  • the plurality of candidate sentence types include non-question sentences and a plurality of question sentence types; as shown in FIG. 6 , the classifier 202 includes a plurality of first probability calculation models 212 corresponding to the plurality of candidate sentence types respectively and the first decision-making unit 222;
  • each first probability calculation model 212 in the plurality of first probability calculation models is used for determining the probability that the sentence type corresponding to the semantic representation information is the candidate sentence type for its corresponding candidate sentence type;
  • the first decision unit 222 is configured to determine the sentence type corresponding to the semantic representation information according to the probability corresponding to each candidate sentence type.
  • the first decision unit 222 is configured to select the maximum probability from the probabilities corresponding to each candidate sentence type, and determine the sentence type corresponding to the maximum probability as the sentence type corresponding to the semantic representation information.
  • the plurality of question types include open-ended questions, intellectual questions, dialog management questions, and procedural questions.
  • the multiple candidate sentence types include questions and non-questions; as shown in FIG. 7 , the classifier 202 includes a second probability calculation model 232 and a second decision unit 242;
  • the second probability calculation model 232 is used to determine the probability that the sentence type corresponding to the semantic representation information is a question sentence
  • the second decision unit 242 is configured to determine, according to the probability, whether the sentence type corresponding to the semantic representation information is a question sentence or a non-question sentence.
  • the semantic representation information includes multiple dimension information
  • the second probability calculation model 232 is configured to obtain the probability that the sentence type corresponding to the semantic representation information is a question sentence by calculating the inner product of multiple dimensional information and normalizing the inner product.
  • the device further includes:
  • the word segmenter 203 is used to segment the target sentence to obtain at least one word information in the target sentence;
  • the semantic recognition model 201 acquires corresponding semantic representation information according to at least one word information in the target sentence.
  • FIG. 9 shows a structural block diagram of an electronic device according to an embodiment of the present application.
  • the electronic device includes: a memory 910 and a processor 920 , and a computer program that can be executed on the processor 920 is stored in the memory 910 .
  • the processor 920 executes the computer program, the statement type identification method in the above-mentioned embodiment is implemented.
  • the number of the memory 910 and the processor 920 may be one or more.
  • the electronic equipment also includes:
  • the communication interface 930 is used to communicate with external devices and perform data interactive transmission.
  • the bus can be an industry standard architecture (Industry Standard Architecture, ISA) bus, a peripheral device interconnect (Peripheral Component Interconnect, PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus and the like.
  • ISA Industry Standard Architecture
  • PCI peripheral device interconnect
  • EISA Extended Industry Standard Architecture
  • the bus can be divided into address bus, data bus, control bus and so on. For ease of presentation, only one thick line is used in FIG. 9, but it does not mean that there is only one bus or one type of bus.
  • the memory 910, the processor 920 and the communication interface 930 are integrated on one chip, the memory 910, the processor 920 and the communication interface 930 can communicate with each other through an internal interface.
  • the embodiments of the present application provide a computer-readable storage medium, which stores a computer program, and when the program is executed by a processor, implements the methods provided in the embodiments of the present application.
  • An embodiment of the present application also provides a chip, the chip includes, and includes a processor, configured to call and execute an instruction stored in the memory from a memory, so that a communication device installed with the chip executes the method provided by the embodiment of the present application.
  • An embodiment of the present application further provides a chip, including: an input interface, an output interface, a processor, and a memory, the input interface, the output interface, the processor, and the memory are connected through an internal connection path, and the processor is used to execute codes in the memory , when the code is executed, the processor is used to execute the method provided by the embodiment of the application.
  • processor may be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processing (digital signal processing, DSP), application specific integrated circuit (application specific integrated circuit, ASIC), field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general purpose processor may be a microprocessor or any conventional processor or the like. It should be noted that the processor may be a processor supporting an advanced RISC machine (ARM) architecture.
  • ARM advanced RISC machine
  • the above-mentioned memory may include read-only memory and random access memory, and may also include non-volatile random access memory.
  • the memory may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
  • the non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically programmable read-only memory (EPROM). Erase programmable read-only memory (electrically EPROM, EEPROM) or flash memory.
  • Volatile memory may include random access memory (RAM), which acts as an external cache. By way of example and not limitation, many forms of RAM are available.
  • SRAM static RAM
  • DRAM dynamic random access memory
  • SDRAM synchronous dynamic random access memory
  • double data rate synchronous dynamic random access Memory double data date SDRAM, DDR SDRAM
  • enhanced synchronous dynamic random access memory enhanced SDRAM, ESDRAM
  • synchronous link dynamic random access memory direct memory bus random access memory
  • direct rambus RAM direct rambus RAM
  • the above-mentioned embodiments it may be implemented in whole or in part by software, hardware, firmware or any combination thereof.
  • software it can be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the computer program instructions When the computer program instructions are loaded and executed on a computer, the processes or functions according to the present application result in whole or in part.
  • the computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable device.
  • Computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • first and second are only used for descriptive purposes, and should not be construed as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature delimited with “first”, “second” may expressly or implicitly include at least one of that feature.
  • plurality means two or more, unless otherwise expressly and specifically defined.
  • any description of a process or method in the flowcharts or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing a specified logical function or step of the process .
  • the scope of the preferred embodiments of the present application includes alternative implementations in which the functions may be performed out of the order shown or discussed, including performing the functions substantially concurrently or in the reverse order depending upon the functions involved.
  • each functional unit in each embodiment of the present application may be integrated into one processing module, or each unit may exist physically alone, or two or more units may be integrated into one module.
  • the above-mentioned integrated modules can be implemented in the form of hardware, and can also be implemented in the form of software function modules. If the above-mentioned integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.
  • the storage medium may be a read-only memory, a magnetic disk or an optical disk, and the like.

Abstract

一种语句类型识别方法、装置、电子设备和存储介质，该方法包括：从待识别的课堂教学语音中确定目标语句（S101）；根据目标语句，获取对应的语义表征信息（S102）；从多个候选语句类型中选取语义表征信息对应的语句类型（S103）；将语义表征信息对应的语句类型确定为目标语句的语句类型（S104）。能够提升对目标语句的分类准确度，有利于识别出多样化的语句类型。

Description

语句类型识别方法、装置、电子设备和存储介质 技术领域
本申请涉及自然语言处理领域，尤其涉及一种语句类型识别方法、装置、电子设备和存储介质。
背景技术
教师在课堂上会采用各种教学手段，以激发学生的思考和钻研。这些教学手段包括提问、分配小组作业等。其中，提问是教师在课堂教学中关键的教学手段之一。它能够增加师生之间的互动，提高学生的参与度，使师生之间能够进行开放式的讨论，了解对方的思想，还能引发学生的更深层次的思考。
对课堂上教师的提问进行分析，能够极大地帮助老师逐渐提高教学质量。然而，由于课堂上教师的话语量大且类型多样，利用传统的自然语言处理方法无法准确识别教师话语中的语句类型。
发明内容
本申请实施例提供一种语句类型识别方法、装置、电子设备和存储介质，以解决相关技术存在的问题，技术方案如下：
第一方面，本申请实施例提供了一种语句类型的方法，包括：
从待识别的课堂教学语音中确定目标语句；
根据目标语句，获取对应的语义表征信息；
从多个候选语句类型中选取语义表征信息对应的语句类型；
将语义表征信息对应的语句类型确定为目标语句的语句类型。
在一种实施方式中，多个候选语句类型包括非问句和多个问句类型；
从多个候选语句类型中选取语义表征信息对应的语句类型，包括：
针对多个候选语句类型中的每个候选语句类型，确定语义表征信息对应的语句类型是候选语句类型的概率；
根据每个候选语句类型对应的概率，确定语义表征信息对应的语句类型。
在一种实施方式中，根据每个候选语句类型对应的概率，确定语义表征信息对应的语句类型，包括：
从每个候选语句类型对应的概率中，选取最大概率；
将最大概率对应的语句类型，确定为语义表征信息对应的语句类型。
在一种实施方式中，多个问句类型包括开放式问句、求知性问句、对话管理问句和程序性问句。
在一种实施方式中，多个候选语句类型包括问句和非问句；
从多个候选语句类型中选取语义表征信息对应的语句类型，包括：
确定语义表征信息对应的语句类型是问句的概率；
根据概率，确定语义表征信息对应的语句类型为问句或非问句。
在一种实施方式中，语义表征信息包括多个维度信息，确定语义表征信息对应的语句类型是问句的概率，包括：
通过计算多个维度信息的内积以及对内积进行归一化，得到语义表征信息对应的语句类型是问句的概率。
在一种实施方式中，该方法还包括：
对目标语句进行分词，得到每个目标语句对应的至少一个单词信息，作为目标语句的特征信息；
根据目标语句，获取对应的语义表征信息，包括：
根据目标语句中的至少一个词信息，获取对应的语义表征信息。
第二方面，本申请实施例提供了一种语句类型识别装置，包括：
语义识别模型，用于根据从待识别的课堂教学语音中确定的目标语句，获取对应的语义表征信息；
分类器，用于从多个候选语句类型中选取语义表征信息对应的语句类型，作为目标语句的语句类型。
在一种实施方式中，多个候选语句类型包括非问句和多个问句类型；分类器包括与多个候选语句类型分别对应的多个第一概率计算模型和第一决策单元；
其中，多个第一概率计算模型中的每个第一概率计算模型，用于针对其对应的候选语句类型，确定语义表征信息对应的语句类型是候选语句类型的概率；
第一决策单元，用于根据多个候选语句中的每个候选语句类型对应的概率，确定语义表征信息对应的语句类型。
在一种实施方式中，第一决策单元用于从多个第一概率计算模型分别确定的概率中选取最大概率，将最大概率对应的语句类型确定为语义表征信息对应的语句类型。
在一种实施方式中，多个问句类型包括开放式问句、求知性问句、对话管理问句和程序性问句。
在一种实施方式中，多个候选语句类型包括问句和非问句；分类器包括第二概率计算模型和第二决策单元；
第二概率计算模型，用于确定语义表征信息对应的语句类型是问句的概率；
第二决策单元，用于根据概率，确定语义表征信息对应的语句类型为问句或非问句。
在一种实施方式中，语义表征信息包括多个维度信息；
第二概率计算模型，用于通过计算多个维度信息的内积以及对内积进行归一化，得到语义表征信息对应的语句类型是问句的概率。
在一种实施方式中，该装置还包括：
分词器，用于对目标语句进行分词，得到目标语句中的至少一个词信息；
语义识别模型根据目标语句中的至少一个词信息，获取对应的语义表征信息。
第三方面，本申请实施例提供了一种电子设备，该电子设备包括：存储器和处理器。其中，该存储器和该处理器通过内部连接通路互相通信，该存储器用于存储指令，该处理器用于执行该存储器存储的指令，并且当该处理器执行该存储器存储的指令时，使得该处理器执行上述各方面任一种实施方式中的方法。
第四方面，本申请实施例提供了一种计算机可读存储介质，计算机可读存储介质存储计算机程序，当计算机程序在计算机上运行时，上述各方面任一种实施方式中的方法被执行。
上述技术方案中的优点或有益效果至少包括：
由于本申请实施例对于从课堂教学语音中确定出的目标语句，先确定语义特征信息，再基于语义表征信息选取对应的语句类型，因此，本申请实施例是基于语义理解对目标语句进行分类，能够提升对目标语句的分类准确度，有利于识别出多样化的语句类型，帮助提升教学质量。
上述概述仅仅是为了说明书的目的，并不意图以任何方式进行限制。除上述描述的示意性的方面、实施方式和特征之外，通过参考附图和以下的详细描述，本申请进一步的方面、实施方式和特征将会是容易明白的。
附图说明
在附图中，除非另外规定，否则贯穿多个附图相同的附图标记表示相同或相似的部件或元素。这些附图不一定是按照比例绘制的。应该理解，这些附图仅描绘了根据本申请公开的一些实施方式，而不应将其视为是对本申请范围的限制。
图1为根据本申请一实施例的语句类型识别方法的示意图。
图2为根据本申请一实施例对课堂教学语音的处理过程的示意图。
图3为根据本申请一实施例的语句类型识别方法的示意图。
图4为根据本申请一实施例的语句类型识别方法的示意图。
图5为根据本申请一实施例的语句类型识别装置的示意图。
图6为根据本申请一实施例的语句类型识别装置的示意图。
图7为根据本申请一实施例的语句类型识别装置的示意图。
图8为根据本申请一实施例的语句类型识别装置的示意图。
图9为根据本申请一实施例的电子设备的结构框图。
具体实施方式
在下文中，仅简单地描述了某些示例性实施例。正如本领域技术人员可认识到的那样，在不脱离本申请的精神或范围的情况下，可通过各种不同方式修改所描述的实施例。因此，附图和描述被认为本质上是示例性的而非限制性的。
图1示出根据本申请一实施例的语句类型识别方法的示意图。如图1所示，该方法可以包括：
步骤S101，从待识别的课堂教学语音中确定目标语句；
步骤S102，根据目标语句，获取对应的语义表征信息；
步骤S103，从多个候选语句类型中选取语义表征信息对应的语句类型；
步骤S104，将语义表征信息对应的语句类型确定为目标语句的语句类型。
在实际应用时，可以将课堂教学语音中的全部语句或部分语句确定为目标语句。目标语句的数量可以是一个或多个。对于多个目标语句中的每个目标语句，分别执行上述步骤S102至S104，得到与每个目标语句分别对应的语句类型。
示例性地，目标语句可以是文本格式的语句。可以对音频格式的课堂教学语音进行文本识别，得到文本格式的目标语句。
例如，如图2所示，对一段课堂教学语音，采用VAD(Voice Activity Detection，语音活动检测)技术，提取出包含人声的音频片段集合。然后，对音频片段集合中的每个音频片段，采用ASR(Voice Activity Detection，语音活动检测)技术，转换得到文本格式的语句集合，对语句集合进行语句划分和选取等操作，得到目标语句。然后，将目标语句输入到可以执行上述步骤S102至S103的语句类型识别装置，则可得到目标语句的语句类型例如类型1、类型2或类型3。
示例性地，该语句类型识别模型可包括语义识别模型和分类器，其中，语义识别模型可根据输入的目标语句或目标语句的特征信息，输出对应的语义表征信息。将多个候选语句类型作为分类器的分类结果，将语义表征信息输入分类器，可利用分类器输出多个候选语句类型中的一个，作为语义表征向量对应的语句类型。
作为示例，语义识别模型可以包括Tranformer模型、BiLSTM(Bi-directional Long Short-Term Memory，双向长短期记忆)模型或BiGRU(bidirection gated recurrent unit，双向门控循环单元)等。分类器可包括一个或多个MLP(Multilayer Perceptron，多层感知机)或CNN(Convolutional Neural Network，卷积神经网络)。
在一种示例性的实施方式中，在步骤S102之前，语句类型的方法还可以包括：
对目标语句进行分词，得到目标语句中的至少一个词信息；
相应地，步骤S102中，根据目标语句，获取对应的语义表征信息，包 括：根据目标语句中的至少一个词信息，获取对应的语义表征信息。
例如，将目标语句分为P个词语，将每一个词映射为对应的词向量，得到P个词向量的集合(X 1,X 2,…,X P)，其中，每个词向量可以是多维向量，例如是一个200维向量X I＝(E 1,E 2,…,E 200)。词向量例如是one-hot(独热)向量或分布式词向量等。
根据该示例性的实施方式，先将目标语句映射为词向量，实现用机器语言表征目标语句，有利于准确获取目标语句的特征信息。
示例性地，语义表征信息也可包括多个维度信息。即语义表征信息也可以用多维向量表示。
举例而言，如图3所示，将目标语句的每个N维词向量E [CLS]以及对应的维度信息E 1至E N，输入到语义识别模型如Transformer(Trm)模型中。Transformer模型对各词向量进行处理，输出语义表征向量H [CLS]。其中，语义表征向量可包括K个维度信息H 1至H K，例如K＝128，语义表征向量为128维向量。将语义表征向量输入至分类器，利用分类器可输出语义表征向量对应的语句类型，作为目标语句的语句类型。
由于本申请实施例对于从课堂教学语音中确定出的目标语句，先确定语义特征信息，再基于语义表征信息选取对应的语句类型，因此，本申请实施例是基于语义理解对目标语句进行分类，能够提升对目标语句的分类准确度，有利于识别出多样化的语句类型。
利用本申请实施例的方法，不仅能够识别出问句，还能够识别出具体的问句类型。例如，可将问句分为以下多种类型：
1、开放问句(Open Question)：引发思考类的问句例如与教学内容相关的没有标准答案的问句。例如让学生讲题、讲知识点或者总结归纳、分析或推测的提问。如：“这道题怎么做？”、“这道题你为什么错了？”、“你可以讲一下你的思路吗？”、“所以说你应该最后结论是什么？”、“是一个什么样的事什么样的原理呢？”、“第二条路该怎么做呢？”、“这个这种结构应该是什么样子？”等。
2、求知性问句(Knowledge-solicitation Question)：例如与教学内容相关的具有标准答案的问句。如：“它的距离是多少？”、“那他的施受力物体分别是什么？”、“这道题答案是啥？”、“四边形是什么呢？”
3、对话管理问句(Discourse-management Question)：例如与教学内容无关的主要用于引起注意的问句，包括“是吧？”、“对吧？”、“对不对？”、“是不是？”等。例如，“对，前面讲那个题也是这样吧？”、“可以的，对不对？”或者“因为什么呢？因为……啊”等不是必须回答的问句。
4、程序性问句(Procedural Question)：例如与教学内容无关但与教学程序相关的问句，例如读题目、调试设备、和学生打招呼、询问最近生活情况等的问句。例如：“最近如何？”、“能听见么？”、“你还有问题呢？咱们先把问题问完，我再给你讲这些”、“我是不是跟你讲过？”、“讲到哪儿？”等。
通过将问句分为上述多种类型，并利用本申请实施例提供的方法识别课堂教学语音中的目标语句的语句类型，可以帮助评估和改善教学质量。例如，通过统计开放性问句和求知性问句的占比，可以评估教师是否善于引导学生思考。又如，通过识别出求知性问句，确定求知性问句后的应答情况，可以了解学生在课堂上对教学内容的掌握程度，有助于教师针对性地调整教学进度。再如，通过统计对话管理问句或程序性问句在所有语句中的占比，可以评估教师的教学效率，了解教学过程是否流畅，以及对教学环境进行改善或提醒教师提高效率。
在一种实施方式中，上述步骤S103中的多个候选语句类型包括非问句和多个问句类型。从多个候选语句类型中选取语义表征信息对应的语句类型可以包括：
针对多个候选语句类型中的每个候选语句类型，确定语义表征信息对应的语句类型是候选语句类型的概率；
根据每个候选语句类型对应的概率，确定语义表征信息对应的语句类型。
举例而言，参考图4，将语义表征向量H [CLS]输入分类器中，分类器中包括多个概率计算模型例如图中的MLP。或者，也可以用DNN模型，例如包括大小分别为128、64、32的三个卷积层的DNN模型。每个概率计算模型对应于一种候选语句类型，包括非问句Y 1和多个问句类型Y 2至Y M。每个MLP计算出语义表征向量H [CLS]对应的语句类型为MLP对应的Y的概率，即可利用各候选语句类型Y 1至Y M对应的概率，确定出语义表征信息对应的语句类型。
示例性地，在确定语义表征信息对应的语句类型之前，可以对多个候选语句类型对应的概率进行归一化，以将多个候选语句类型的概率转换到同一分布空间中，提高确定语句类型的准确性。例如利用softmax对多个概率进行归一化。
示例性地，根据每个候选语句类型对应的概率，确定语义表征信息对应的语句类型，可以包括：
从每个候选语句类型对应的概率中，选取最大概率；
将最大概率对应的语句类型，确定为语义表征信息对应的语句类型。
根据上述实施方式，对于不同的候选问句类型，可获取不同的概率计算模型，分别计算其对应的概率。因此，可以利用更具针对性的训练数据或采用更具针对性的网络结构，得到准确度更高的概率计算模型，进一步提高语句类型识别准确度。
在一种实施方式中，候选语句也可以包括问句和非问句。在这种情况下，上述步骤S103，从多个候选语句类型中选取语义表征信息对应的语句类型，包括：
确定语义表征信息对应的语句类型是问句的概率；
根据概率，确定语义表征信息对应的语句类型为问句或非问句。
例如，利用一个概率计算模型确定语义表征信息对应的语句类型是问句的概率是0.6，该概率大于阈值0.5，则确定该语义表征信息对应的语句类型为问句。具体实施时，根据实际应用场景的需求例如需提高识别出的问句的准确度或需识别出全部问句，将该阈值调高或调低。
示例性地，可以通过计算多个维度信息的内积以及对内积进行归一化，得到语义表征信息对应的语句类型是问句的概率。
由于本申请实施例基于语义表征信息确定语句类型，在一定程度上提升了语句类型识别准确度，因此，在仅需识别问句和非问句的场合，通过计算内积将多个维度信息综合汇总，然后进行归一化，能够获得准确度符合要求的识别结果并降低计算复杂度。
综上，由于本申请实施例对于从课堂教学语音中确定出的目标语句，先确定语义特征信息，再基于语义表征信息选取对应的语句类型，因此，本申请实施例是基于语义理解对目标语句进行分类，能够提升对目标语句的分类准确度，有利于识别出多样化的语句类型，帮助提升教学质量。
图5示出根据本申请一实施例的语句类型识别装置的示意图。该装置可以是用于执行上述语句类型识别方法的电子设备，也可以是语句类型识别模型，例如图3和图4所示的模型。
如图5所示，该装置可以包括：
语义识别模型201，用于根据从待识别的课堂教学语音中确定的目标语句，获取对应的语义表征信息；
分类器202，用于从多个候选语句类型中选取语义表征信息对应的语句类型，作为目标语句的语句类型。
在一种实施方式中，多个候选语句类型包括非问句和多个问句类型；如图6所示，分类器202包括与多个候选语句类型分别对应的多个第一概率计算模型212和第一决策单元222；
其中，多个第一概率计算模型中的每个第一概率计算模型212，用于针对其对应的候选语句类型，确定语义表征信息对应的语句类型是候选语句类型的概率；
第一决策单元222，用于根据每个候选语句类型对应的概率，确定语义表征信息对应的语句类型。
示例性地，第一决策单元222用于从每个候选语句类型对应的概率中选取最大概率，将最大概率对应的语句类型确定为语义表征信息对应的语句类型。
示例性地，多个问句类型包括开放式问句、求知性问句、对话管理问句和程序性问句。
在一种实施方式中，多个候选语句类型包括问句和非问句；如图7所示，分类器202包括第二概率计算模型232和第二决策单元242；
第二概率计算模型232，用于确定语义表征信息对应的语句类型是问句的概率；
第二决策单元242，用于根据概率，确定语义表征信息对应的语句类型为问句或非问句。
示例性地，语义表征信息包括多个维度信息；
第二概率计算模型232用于通过计算多个维度信息的内积以及对内积进行归一化，得到语义表征信息对应的语句类型是问句的概率。
在一种实施方式中，如图8所示，该装置还包括：
分词器203，用于对目标语句进行分词，得到目标语句中的至少一个词信息；
语义识别模型201根据目标语句中的至少一个词信息，获取对应的语义表征信息。
本申请实施例各装置中的各模块的功能可以参见上述方法中的对应描述，在此不再赘述。
图9示出根据本申请一实施例的电子设备的结构框图。如图9所示，该电子设备包括：存储器910和处理器920，存储器910内存储有可在处理器920上运行的计算机程序。处理器920执行该计算机程序时实现上述实施例中的语句类型识别方法。存储器910和处理器920的数量可以为一个或多个。
该电子设备还包括：
通信接口930，用于与外界设备进行通信，进行数据交互传输。
如果存储器910、处理器920和通信接口930独立实现，则存储器910、处理器920和通信接口930可以通过总线相互连接并完成相互间的通信。该总线可以是工业标准体系结构(Industry Standard Architecture，ISA)总线、外部设备互连(Peripheral Component Interconnect，PCI)总线或扩展工业标准体系结构(Extended Industry Standard Architecture，EISA)总线等。该总线可以分为地址总线、数据总线、控制总线等。为便于表示，图9中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
可选的，在具体实现上，如果存储器910、处理器920及通信接口930集成在一块芯片上，则存储器910、处理器920及通信接口930可以通过内部接口完成相互间的通信。
本申请实施例提供了一种计算机可读存储介质，其存储有计算机程序，该程序被处理器执行时实现本申请实施例中提供的方法。
本申请实施例还提供了一种芯片，该芯片包括，包括处理器，用于从存储器中调用并运行存储器中存储的指令，使得安装有芯片的通信设备执行本申请实施例提供的方法。
本申请实施例还提供了一种芯片，包括：输入接口、输出接口、处理器和存储器，输入接口、输出接口、处理器以及存储器之间通过内部连接通 路相连，处理器用于执行存储器中的代码，当代码被执行时，处理器用于执行申请实施例提供的方法。
应理解的是，上述处理器可以是中央处理器(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(digital signal processing，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现场可编程门阵列(fieldprogrammablegate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者是任何常规的处理器等。值得说明的是，处理器可以是支持进阶精简指令集机器(advanced RISC machines，ARM)架构的处理器。
进一步地，可选的，上述存储器可以包括只读存储器和随机存取存储器，还可以包括非易失性随机存取存储器。该存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以包括只读存储器(read-only memory，ROM)、可编程只读存储器(programmable ROM，PROM)、可擦除可编程只读存储器(erasable PROM，EPROM)、电可擦除可编程只读存储器(electrically EPROM，EEPROM)或闪存。易失性存储器可以包括随机存取存储器(random access memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用。例如，静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(dynamic random access memory，DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data date SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM，SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM，DR RAM)。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行计算机程序指令时，全部或部分地产生按照本申请的流程或功能。计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输。
在本说明书的描述中，参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包括于本申请的至少一个实施例或示例中。而且，描述的具体特征、结构、材料或者特点可以在任一个或多个实施例或示例中以合适的方式结合。此外，在不相互矛盾的情况下，本领域的技术人员可以将本说明书中描述的不同实施例或示例以及不同实施例或示例的特征进行结合和组合。
此外，术语“第一”、“第二”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或隐含地包括至少一个该特征。在本申请的描述中，“多个”的含义是两个或两个以上，除非另有明确具体的限定。
流程图中或在此以其他方式描述的任何过程或方法描述可以被理解为，表示包括一个或更多个用于实现特定逻辑功能或过程的步骤的可执行指令的代码的模块、片段或部分。并且本申请的优选实施方式的范围包括另外的实现，其中可以不按所示出或讨论的顺序，包括根据所涉及的功能按基本同时的方式或按相反的顺序，来执行功能。
在流程图中表示或在此以其他方式描述的逻辑和/或步骤，例如，可以被认为是用于实现逻辑功能的可执行指令的定序列表，可以具体实现在任何计算机可读介质中，以供指令执行系统、装置或设备(如基于计算机的系统、包括处理器的系统或其他可以从指令执行系统、装置或设备取指令并执行指令的系统)使用，或结合这些指令执行系统、装置或设备而使用。
应理解的是，本申请的各部分可以用硬件、软件、固件或它们的组合来实现。在上述实施方式中，多个步骤或方法可以用存储在存储器中且由合适的指令执行系统执行的软件或固件来实现。上述实施例方法的全部或部分步骤是可以通过程序来指令相关的硬件完成，该程序可以存储于一种计算机可读存储介质中，该程序在执行时，包括方法实施例的步骤之一或其组合。
此外，在本申请各个实施例中的各功能单元可以集成在一个处理模块中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个模块中。上述集成的模块既可以采用硬件的形式实现，也可以采用软件功能模块的形式实现。上述集成的模块如果以软件功能模块的形式实现并作为独立的产品销售或使用时，也可以存储在一个计算机可读存储介质中。该存储介质可以是只读存储器，磁盘或光盘等。
以上，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到其各种变化或替换，这些都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (16)

  1. 一种语句类型识别方法，其特征在于，包括：
    从待识别的课堂教学语音中确定目标语句；
    根据所述目标语句，获取对应的语义表征信息；
    从多个候选语句类型中选取所述语义表征信息对应的语句类型；
    将所述语义表征信息对应的语句类型确定为所述目标语句的语句类型。
  2. 根据权利要求1所述的方法，其特征在于，所述多个候选语句类型包括非问句和多个问句类型；
    所述从多个候选语句类型中选取所述语义表征信息对应的语句类型，包括：
    针对所述多个候选语句类型中的每个候选语句类型，确定所述语义表征信息对应的语句类型是所述候选语句类型的概率；
    根据所述每个候选语句类型对应的概率，确定所述语义表征信息对应的语句类型。
  3. 根据权利要求2所述的方法，其特征在于，根据所述每个候选语句类型对应的概率，确定所述语义表征信息对应的语句类型，包括：
    从所述每个候选语句类型对应的概率中，选取最大概率；
    将所述最大概率对应的语句类型，确定为所述语义表征信息对应的语句类型。
  4. 根据权利要求3所述的方法，其特征在于，所述多个问句类型包括开放式问句、求知性问句、对话管理问句和程序性问句。
  5. 根据权利要求2所述的方法，其特征在于，所述多个候选语句类型包括问句和非问句；
    所述从多个候选语句类型中选取所述语义表征信息对应的语句类型，包括：
    确定所述语义表征信息对应的语句类型是问句的概率；
    根据所述概率，确定所述语义表征信息对应的语句类型为问句或非问句。
  6. 根据权利要求5所述的方法，其特征在于，所述语义表征信息包括多个维度信息，所述确定所述语义表征信息对应的语句类型是问句的概率，包括：
    通过计算所述多个维度信息的内积以及对所述内积进行归一化，得到所述语义表征信息对应的语句类型是问句的概率。
  7. 根据权利要求1至6中任一项所述的方法，其特征在于，所述方法还包括：
    对所述目标语句进行分词，得到所述目标语句中的至少一个词信息；
    所述根据所述目标语句，获取对应的语义表征信息，包括：
    根据所述目标语句中的至少一个词信息，获取对应的语义表征信息。
  8. 一种语句类型识别装置，其特征在于，包括：
    语义识别模型，用于根据从待识别的课堂教学语音中确定的目标语句，获取对应的语义表征信息；
    分类器，用于从多个候选语句类型中选取所述语义表征信息对应的语句类型，作为所述目标语句的语句类型。
  9. 根据权利要求8所述的装置，其特征在于，所述多个候选语句类型包括非问句和多个问句类型；所述分类器包括与所述多个候选语句类型分别对应的多个第一概率计算模型和第一决策单元；
    其中，所述多个第一概率计算模型中的每个第一概率计算模型，用于针对其对应的候选语句类型，确定所述语义表征信息对应的语句类型是所述候选语句类型的概率；
    所述第一决策单元，用于根据所述多个候选语句类型中的每个候选语句类型对应的概率，确定所述语义表征信息对应的语句类型。
  10. 根据权利要求9所述的装置，其特征在于，所述第一决策单元用于从所述每个候选语句类型对应的概率中选取最大概率，将所述最大概率对应的语句类型确定为所述语义表征信息对应的语句类型。
  11. 根据权利要求9所述的装置，其特征在于，所述多个问句类型包括开放式问句、求知性问句、对话管理问句和程序性问句。
  12. 根据权利要求8所述的装置，其特征在于，所述多个候选语句类型包括问句和非问句；所述分类器包括第二概率计算模型和第二决策单元；
    所述第二概率计算模型，用于确定所述语义表征信息对应的语句类型是问句的概率；
    所述第二决策单元，用于根据所述概率，确定所述语义表征信息对应的语句类型为问句或非问句。
  13. 根据权利要求12所述的装置，其特征在于，所述语义表征信息包括多个维度信息；
    所述第二概率计算模型，用于通过计算所述多个维度信息的内积以及对所述内积进行归一化，得到所述语义表征信息对应的语句类型是问句的概率。
  14. 根据权利要求8至13中任一项所述的装置，其特征在于，所述装置还包括：
    分词器，用于对所述目标语句进行分词，得到所述目标语句中的至少一个词信息；
    所述语义识别模型根据所述目标语句中的至少一个词信息，获取对应的语义表征信息。
  15. 一种电子设备，其特征在于，包括：处理器和存储器，所述存储器中存储指令，所述指令由处理器加载并执行，以实现如权利要求1至7中 任一项所述的方法。
  16. 一种计算机可读存储介质，所述计算机可读存储介质内存储有计算机程序，所述计算机程序被处理器执行时实现如权利要求1至7中任一项所述的方法。
PCT/CN2021/104353 2020-07-06 2021-07-02 语句类型识别方法、装置、电子设备和存储介质 WO2022007723A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/147,080 US11775769B2 (en) 2020-07-06 2022-12-28 Sentence type recognition method and apparatus, electronic device, and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010642759.5 2020-07-06
CN202010642759.5A CN111753553B (zh) 2020-07-06 2020-07-06 语句类型识别方法、装置、电子设备和存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/147,080 Continuation US11775769B2 (en) 2020-07-06 2022-12-28 Sentence type recognition method and apparatus, electronic device, and storage medium

Publications (1)

Publication Number Publication Date
WO2022007723A1 true WO2022007723A1 (zh) 2022-01-13

Family

ID=72679547

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/104353 WO2022007723A1 (zh) 2020-07-06 2021-07-02 语句类型识别方法、装置、电子设备和存储介质

Country Status (3)

Country Link
US (1) US11775769B2 (zh)
CN (1) CN111753553B (zh)
WO (1) WO2022007723A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116050412A (zh) * 2023-03-07 2023-05-02 江西风向标智能科技有限公司 基于数学语义逻辑关系的高中数学题目的分割方法和系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111753553B (zh) 2020-07-06 2022-07-05 北京世纪好未来教育科技有限公司 语句类型识别方法、装置、电子设备和存储介质
CN116597461B (zh) * 2023-07-14 2023-09-22 广东信聚丰科技股份有限公司 基于人工智能的题目知识点关联方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102231278A (zh) * 2011-06-10 2011-11-02 安徽科大讯飞信息科技股份有限公司 实现语音识别中自动添加标点符号的方法及系统
US20130018649A1 (en) * 2011-07-13 2013-01-17 Nuance Communications, Inc. System and a Method for Generating Semantically Similar Sentences for Building a Robust SLM
CN107357781A (zh) * 2017-06-29 2017-11-17 胡玥莹 用于鉴别网页标题与正文关联度的系统及方法
CN108415980A (zh) * 2018-02-09 2018-08-17 平安科技（深圳）有限公司 问答数据处理方法、电子装置及存储介质
CN111753553A (zh) * 2020-07-06 2020-10-09 北京世纪好未来教育科技有限公司 语句类型识别方法、装置、电子设备和存储介质

Family Cites Families (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01314373A (ja) * 1988-06-15 1989-12-19 Hitachi Ltd 機械翻訳システムにおける訳語選択方式
CN105701120B (zh) * 2014-11-28 2019-05-03 华为技术有限公司 确定语义匹配度的方法和装置
CN105261362B (zh) 2015-09-07 2019-07-05 科大讯飞股份有限公司 一种通话语音监测方法及系统
CN106528530A (zh) * 2016-10-24 2017-03-22 北京光年无限科技有限公司 一种确定句子类型的方法及装置
CN107330100B (zh) * 2017-07-06 2020-04-03 北京大学深圳研究生院 基于多视图联合嵌入空间的图像-文本双向检索方法
CN110019722B (zh) * 2017-12-21 2023-11-24 株式会社理光 对话模型的回复排序方法、装置及计算机可读存储介质
CN110502738A (zh) * 2018-05-18 2019-11-26 阿里巴巴集团控股有限公司 中文命名实体识别方法、装置、设备和查询系统
US11074253B2 (en) * 2018-11-02 2021-07-27 International Business Machines Corporation Method and system for supporting inductive reasoning queries over multi-modal data from relational databases
CN109408627B (zh) * 2018-11-15 2021-03-02 众安信息技术服务有限公司 一种融合卷积神经网络和循环神经网络的问答方法及系统
CN109918376A (zh) 2019-02-26 2019-06-21 北京致远互联软件股份有限公司 数据表处理方法、装置以及电子设备
CN109918673B (zh) * 2019-03-14 2021-08-03 湖北亿咖通科技有限公司 语义仲裁方法、装置、电子设备和计算机可读存储介质
US11651032B2 (en) * 2019-05-03 2023-05-16 Servicenow, Inc. Determining semantic content of textual clusters
CN110321419B (zh) * 2019-06-28 2021-06-15 神思电子技术股份有限公司 一种融合深度表示与交互模型的问答匹配方法
CN110619051B (zh) * 2019-08-16 2023-08-04 科大讯飞(苏州)科技有限公司 问题语句分类方法、装置、电子设备及存储介质
CN111125331B (zh) * 2019-12-20 2023-10-31 京东方科技集团股份有限公司 语义识别方法、装置、电子设备及计算机可读存储介质
WO2021251953A1 (en) * 2020-06-09 2021-12-16 Google Llc Generation of interactive audio tracks from visual content
US11335062B2 (en) * 2020-08-14 2022-05-17 Accenture Global Solutions Limited Automated apparel design using machine learning
US20220093088A1 (en) * 2020-09-24 2022-03-24 Apple Inc. Contextual sentence embeddings for natural language processing applications
US11734025B2 (en) * 2020-10-14 2023-08-22 Servicenow, Inc. Configurable action generation for a remote network management platform
TWI778442B (zh) * 2020-11-03 2022-09-21 財團法人資訊工業策進會 偵測文章目的之裝置及方法
US11783008B2 (en) * 2020-11-06 2023-10-10 Adobe Inc. Machine-learning tool for generating segmentation and topic metadata for documents
US11907678B2 (en) * 2020-11-10 2024-02-20 International Business Machines Corporation Context-aware machine language identification
US11264140B1 (en) * 2020-12-16 2022-03-01 Ro5 Inc. System and method for automated pharmaceutical research utilizing context workspaces

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102231278A (zh) * 2011-06-10 2011-11-02 安徽科大讯飞信息科技股份有限公司 实现语音识别中自动添加标点符号的方法及系统
US20130018649A1 (en) * 2011-07-13 2013-01-17 Nuance Communications, Inc. System and a Method for Generating Semantically Similar Sentences for Building a Robust SLM
CN107357781A (zh) * 2017-06-29 2017-11-17 胡玥莹 用于鉴别网页标题与正文关联度的系统及方法
CN108415980A (zh) * 2018-02-09 2018-08-17 平安科技（深圳）有限公司 问答数据处理方法、电子装置及存储介质
CN111753553A (zh) * 2020-07-06 2020-10-09 北京世纪好未来教育科技有限公司 语句类型识别方法、装置、电子设备和存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116050412A (zh) * 2023-03-07 2023-05-02 江西风向标智能科技有限公司 基于数学语义逻辑关系的高中数学题目的分割方法和系统
CN116050412B (zh) * 2023-03-07 2024-01-26 江西风向标智能科技有限公司 基于数学语义逻辑关系的高中数学题目的分割方法和系统

Also Published As

Publication number Publication date
US20230134101A1 (en) 2023-05-04
CN111753553A (zh) 2020-10-09
US11775769B2 (en) 2023-10-03
CN111753553B (zh) 2022-07-05

Similar Documents

Publication Publication Date Title
WO2022007723A1 (zh) 语句类型识别方法、装置、电子设备和存储介质
WO2020237869A1 (zh) 一种问题意图识别方法、装置、计算机设备及存储介质
CN107590127B (zh) 一种题库知识点自动标注方法及系统
KR20190125154A (ko) 심리 상담 데이터를 기계 학습한 자동 대화 장치 및 그 방법
US20230080671A1 (en) User intention recognition method and apparatus based on statement context relationship prediction
CN108536670A (zh) 输出语句生成装置、方法和程序
KR102024845B1 (ko) 화행 분석 장치 및 방법
CN117149989B (zh) 大语言模型训练方法、文本处理方法及装置
US11907665B2 (en) Method and system for processing user inputs using natural language processing
US11615787B2 (en) Dialogue system and method of controlling the same
CN112786028A (zh) 声学模型处理方法、装置、设备和可读存储介质
CN116597809A (zh) 多音字消歧方法、装置、电子设备及可读存储介质
CN115081452B (zh) 一种实体关系的抽取方法
KR20210059995A (ko) 학습 기반의 외국어 말하기 평가 방법 및 그 시스템
CN114242045A (zh) 一种自然语言对话系统意图深度学习方法
CN113806475A (zh) 信息回复方法、装置、电子设备和存储介质
Senevirathne et al. Smart Personal Intelligent Assistant for Candidates of IELTS Exams
CN113012685A (zh) 音频识别方法、装置、电子设备及存储介质
KR20200072005A (ko) 음성 인식된 문장의 보정 방법
CN115083437B (zh) 一种确定学习者发音的不确定性的方法及装置
CN117711444B (zh) 一种基于口才表达的互动方法、装置、设备及存储介质
KR102556797B1 (ko) 인공지능 모델 기반 외국어 말하기 능력에 대한 자동 평가 및 사용자 맞춤형 외국어 교육 콘텐츠 큐레이션 서비스 제공 방법, 장치 및 시스템
CN113555006B (zh) 一种语音信息识别方法、装置、电子设备及存储介质
US11521283B2 (en) Assigning a student to a cohort on a platform
Zhang et al. Optimization and Evaluation of Spoken English CAF Based on Artificial Intelligence and Corpus

Legal Events

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

Ref document number: 21838938

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

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

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 250423)

122 Ep: pct application non-entry in european phase

Ref document number: 21838938

Country of ref document: EP

Kind code of ref document: A1