WO2021092730A1 - 摘要生成方法、装置、电子设备和存储介质 - Google Patents

摘要生成方法、装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2021092730A1
WO2021092730A1 PCT/CN2019/117230 CN2019117230W WO2021092730A1 WO 2021092730 A1 WO2021092730 A1 WO 2021092730A1 CN 2019117230 W CN2019117230 W CN 2019117230W WO 2021092730 A1 WO2021092730 A1 WO 2021092730A1
Authority
WO
WIPO (PCT)
Prior art keywords
abstract
text
similarity
degree
translation
Prior art date
Application number
PCT/CN2019/117230
Other languages
English (en)
French (fr)
Inventor
郝杰
Original Assignee
深圳市欢太科技有限公司
Oppo广东移动通信有限公司
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 深圳市欢太科技有限公司, Oppo广东移动通信有限公司 filed Critical 深圳市欢太科技有限公司
Priority to CN201980101002.0A priority Critical patent/CN114514529A/zh
Priority to PCT/CN2019/117230 priority patent/WO2021092730A1/zh
Publication of WO2021092730A1 publication Critical patent/WO2021092730A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions

Definitions

  • This application relates to machine translation technology, in particular to a method, device, electronic device, and storage medium for generating abstracts.
  • embodiments of the present application provide a method, device, electronic device, and storage medium for generating a summary.
  • the embodiment of the present application provides a method for generating a summary, including:
  • the second abstract or the third abstract is selected as the abstract translation of the first text; wherein,
  • the first text, the first abstract, and the fourth abstract correspond to a first language; the second text, the second abstract, and the third abstract correspond to a second language.
  • the selecting the second abstract or the third abstract as the abstract translation of the first text based on the first similarity and the second similarity includes:
  • the second abstract is selected as the abstract translation.
  • the determining the larger value between the first degree of similarity and the second degree of similarity includes:
  • the difference is greater than a set threshold, the larger value between the first degree of similarity and the second degree of similarity is determined.
  • the selecting the second abstract or the third abstract as the abstract translation of the first text based on the first similarity and the second similarity includes:
  • any one of the second abstract and the third abstract is selected as the abstract translation.
  • the method further includes:
  • the determining the first text includes:
  • the determining the second text includes:
  • the determining the second text includes:
  • the method further includes:
  • the abstract translation of the first text is displayed while the speaker is speaking; wherein, the content of the speaker's speech is related to the text content of the first text.
  • the embodiment of the present application also provides a summary generating device, which includes:
  • a generating unit configured to generate a first abstract about the first text, and generate a second abstract about the second text; the second text is a translated text corresponding to the first text;
  • a translation unit configured to translate the first abstract to obtain a third abstract, and to translate the second abstract to obtain a fourth abstract
  • a first determining unit configured to determine a first degree of similarity between the second abstract and the third abstract, and to determine a second degree of similarity between the first abstract and the fourth abstract;
  • the second determining unit is configured to select the second abstract or the third abstract as the abstract translation of the first text based on the first similarity and the second similarity;
  • the first text, the first abstract, and the fourth abstract correspond to a first language; the second text, the second abstract, and the third abstract correspond to a second language.
  • the embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor.
  • the processor executes the computer program, the foregoing Steps of any summary generation method.
  • the embodiment of the present application also provides a storage medium on which computer instructions are stored, and when the computer instructions are executed by a processor, the steps of any one of the above-mentioned abstract generation methods are implemented.
  • the abstract generation method, device, electronic device, and storage medium provided by the embodiments of the present application generate a first abstract about the first text, and generate a second abstract about the translated text corresponding to the first text, and translate the first abstract, Obtain the third abstract and rewind the second abstract to obtain the fourth abstract.
  • the two similarities are compared, and the second abstract or the third abstract is selected as the abstract translation of the first text according to the comparison result, so that the final abstract translation can more accurately reflect the content of the first text .
  • adopting the solutions of the embodiments of the present application can generate abstracts that more accurately reflect the content of the speech, and improve the accuracy of information transmission.
  • Figure 1 is a schematic diagram of the system architecture of the application of the abstract generation method in related technologies
  • FIG. 2 is a schematic flowchart of a method for generating a summary according to an embodiment of the application
  • FIG. 3 is a schematic flowchart of a method for generating a summary according to an embodiment of the application
  • FIG. 4 is a schematic flowchart of a method for generating a summary according to an embodiment of the application
  • FIG. 5 is a schematic diagram of the composition structure of a summary generating device according to an embodiment of the application.
  • FIG. 6 is a schematic diagram of the hardware composition structure of an electronic device according to an embodiment of the application.
  • Figure 1 is a schematic diagram of the system architecture of the application of the abstract generation method in the related art.
  • the system may include: a machine simultaneous interpretation server, a speech recognition server, a translation server, a mobile terminal delivery server, a viewer mobile terminal, a PC (Personal Computer) client, and a display screen.
  • a machine simultaneous interpretation server a speech recognition server
  • a translation server a mobile terminal delivery server
  • a viewer mobile terminal a viewer mobile terminal
  • PC Personal Computer
  • a speaker can speak at a conference through a PC client, and project documents, such as presentation software (PPT, PowerPoint) documents, to the display screen, and show them to the user through the display screen.
  • the PC client collects the speaker’s audio and sends the collected audio to the machine simultaneous interpretation server.
  • the machine simultaneous interpretation server recognizes the audio data through the voice recognition server to obtain the recognized text. Then translate the recognized text through the translation server to obtain the translation result; the machine simultaneous interpretation server sends the translation result to the PC client, and sends the translation result to the viewer's mobile terminal through the mobile terminal delivery server to display the translation for the user
  • the speech content of the speaker can be translated into the language required by the user and displayed.
  • the machine simultaneous interpretation server can also generate a summary translation of the content of the speech based on the translation result of the translation server, and send the summary translation to the audience's mobile terminal through the mobile-side delivery server to help users quickly understand The content of the speaker's statement.
  • each speaker has a longer speech time and more speech content.
  • the translation server performs machine translation on the speech content to obtain the translation There will be more or less translation errors in the results.
  • the generated abstract translation will not accurately reflect the content of the speech, which will have a negative effect on the user's accurate understanding of the content of the speech and reduce the accuracy of information transmission.
  • two different methods are used to obtain two abstract translations about the content of the speech, and the translations are reversed, the similarity between the original abstracts of different abstracts is compared, and the translations of different abstracts are compared.
  • one paragraph is finally selected as the abstract translation about the content of the speech, so that the abstract translation that can more accurately reflect the content of the speech is selected, and the information transmission is accurately realized.
  • FIG. 2 is a schematic flowchart of a method for generating a summary according to an embodiment of the application. As shown in Figure 2, the method includes:
  • Step 201 Generate a first abstract about the first text, and generate a second abstract about the second text; the second text is a translated text corresponding to the first text.
  • the second text is the translated text corresponding to the first text, that is, the first text is the original text, and the second text is the corresponding translation.
  • the first abstract and the second abstract are respectively generated.
  • the methods that can be used can include at least one of the following:
  • the abstract is generated through an abstractive method.
  • generating an abstract in an extractive manner refers to extracting some representative text fragments from the text to form an abstract, and these fragments can be words, sentences, paragraphs, or subsections in the entire text.
  • the fragments in the text are sorted first, the top-ranked fragments are selected, and the redundant information in the selected fragments is removed, and the summary is finally generated.
  • Generating abstracts in a generative way refers to the establishment of abstract semantic representations and the use of natural language generation techniques to form abstracts.
  • the similarity to the text is high, but the readability is poor, and the amount of information is low.
  • the generative abstracts it has high readability and rich information, but it is similar to the text. The similarity is low.
  • the method further includes:
  • the first text and the second text are determined first.
  • the first text is the original text, and the first text may be the speech manuscript of the speaker.
  • the determining the first text includes:
  • the voice of the speaker can be collected through a voice collection module, such as a microphone, to obtain the first voice, and then the first voice is converted into the first text through voice recognition, thereby obtaining the speaker
  • a voice collection module such as a microphone
  • post-processing such as adding punctuation and text smoothing can be performed on the first text.
  • Adding punctuation allows the first text to achieve reasonable sentence segmentation, and the smooth text can adjust the word order of the text recognized by the voice.
  • the second text is a translation corresponding to the first text.
  • the determining the second text includes:
  • the second text is generated based on the first text, and by translating the determined first text, the obtained translation corresponding to the first text is used as the second text.
  • the translation quality of the speech content of the second text is largely related to the accuracy of the description of the speech content of the first text.
  • the determining the second text includes:
  • the voice of the simultaneous interpretation can be collected through a voice collection module, such as a microphone, to obtain the second voice, and then the second voice can be converted into The second text, from which the translation corresponding to the actual content of the speaker is obtained.
  • a voice collection module such as a microphone
  • the translation quality of the speech content of the second text is no longer related to the accuracy of the description of the speech content of the first text, but is related to the translation quality of simultaneous interpretation.
  • post-processing such as adding punctuation and text smoothing to the second text is also required to obtain the second text that conforms to the grammatical expression.
  • the second text can be obtained in a corresponding manner according to the application scenario.
  • the second text can be obtained by translating the first text; for another example, in the case where the speaker speaks impromptually, speech recognition can be performed on the content of simultaneous interpretation , And then determine the second text.
  • Step 202 Translate the first abstract to obtain a third abstract, and translate the second abstract to obtain a fourth abstract.
  • the first abstract is an abstract about the first text, which is the original text of the abstract
  • the second abstract is an abstract about the second text, which is the translation of the abstract. Therefore, the first abstract is translated, and the third abstract is the translation of the abstract about the first text, and the second abstract is translated, and the fourth abstract is the original abstract of the second text.
  • first text, first abstract and fourth abstract all correspond to the first language, that is, the language used by the speaker when speaking.
  • the above second text, second abstract and third abstract all correspond to the second language , Which is the language of translation.
  • machine translation technology can be used to translate from the first language to the second language, and to realize the translation from the second language to the first language.
  • the server may adopt rule-based machine translation technology or adopt corpus-based machine translation technology for translation.
  • Step 203 Determine the first similarity between the second abstract and the third abstract, and determine the second similarity between the first abstract and the fourth abstract.
  • BLEU bilingual translation quality assistance tools
  • ROUGE Recall-oriented Understanding for Gisting Evaluation
  • Step 204 Based on the first degree of similarity and the second degree of similarity, select the second abstract or the third abstract as the abstract translation of the first text.
  • the second abstract or the third abstract is selected as the abstract translation of the first text based on the first similarity and the second similarity ,include:
  • Step 2041 Determine the larger value between the first degree of similarity and the second degree of similarity
  • Step 2042 When the larger value is the first degree of similarity, select the third abstract as the translation of the abstract; when the larger value is the second degree of similarity, select the second The abstract serves as the translation of the abstract.
  • the comparison result is that the first degree of similarity is greater, that is, the similarity between the second abstract and the third abstract of the same translation language (the second language)
  • the degree is high, it means that the quality of the original abstract (first abstract) generated about the first text is high, and the translation quality of the abstract translation (third abstract) based on the original translation of the abstract is also high, and the third abstract
  • the third abstract Reflecting that the content of the first text is more accurate, the third abstract is used as the translation of the abstract of the first text; when the comparison result is that the second similarity is greater, it is the first language of the original language (first language).
  • the similarity between the abstract and the fourth abstract is high, it means that the translation quality of the second text based on the translation of the first text is high, and the quality of the second abstract generated about the second text is also high.
  • the abstract reflects the high accuracy of the content of the first text, so the second abstract is used as the translation of the abstract of the first text.
  • the determining the larger value between the first similarity degree and the second similarity degree includes:
  • the difference is greater than a set threshold, the larger value between the first degree of similarity and the second degree of similarity is determined.
  • the selecting the second abstract or the third abstract as the abstract translation of the first text based on the first similarity and the second similarity includes:
  • Step 2043 Determine the difference between the first degree of similarity and the second degree of similarity
  • Step 2044 When the difference is less than or equal to a set threshold, select any one of the second abstract and the third abstract as the abstract translation.
  • the method further includes:
  • the abstract translation of the first text is displayed while the speaker is speaking; wherein, the content of the speaker's speech is related to the text content of the first text.
  • the text content of the first text is the speech content of the speaker.
  • the first text can be the speech manuscript of the speaker, or it can be recognized text obtained by real-time voice collection of the speaker’s audio and then voice recognition. . While the speaker is speaking, the abstract translation of the first text is generated. In this way, it can better help participants in the meeting to understand the content of the speaker in real time and grasp the main points of the speech, thereby achieving accurate and rapid information transfer.
  • the abstract generation method provided by the embodiment of the application generates a first abstract about the first text, and generates a second abstract about the translated text corresponding to the first text, translates the first abstract to obtain the third abstract, and compares the first abstract
  • the second abstract is turned back and the fourth abstract is obtained. Then, based on the first similarity between the second abstract and the third abstract, and the second similarity between the first abstract and the fourth abstract, the two similarities The comparison is performed, and the second abstract or the third abstract is selected as the abstract translation of the first text according to the comparison result, so that the final abstract translation can more accurately reflect the content of the first text.
  • adopting the solutions of the embodiments of the present application can generate abstracts that more accurately reflect the content of the speech, and improve the accuracy of information transmission.
  • the solution of the embodiment of the present application can avoid information transmission errors caused by translation errors of machine translation to a certain extent, and more accurately generate a summary of the content of the speech.
  • FIG. 5 is a schematic diagram of the composition structure of a summary generating apparatus according to an embodiment of the application.
  • the summary generating device includes:
  • the generating unit 51 is configured to generate a first abstract about the first text and generate a second abstract about the second text; the second text is a translated text corresponding to the first text;
  • the translation unit 52 is configured to translate the first abstract to obtain a third abstract, and to translate the second abstract to obtain a fourth abstract;
  • the first determining unit 53 is configured to determine a first similarity between the second abstract and the third abstract, and to determine a second similarity between the first abstract and the fourth abstract;
  • the second determining unit 54 is configured to select the second abstract or the third abstract as the abstract translation of the first text based on the first similarity and the second similarity; wherein,
  • the first text, the first abstract, and the fourth abstract correspond to a first language; the second text, the second abstract, and the third abstract correspond to a second language.
  • the second determining unit 54 is further configured to:
  • the second abstract is selected as the abstract translation.
  • the second determining unit 54 determines the larger value between the first similarity and the second similarity, it is further configured to:
  • the difference is greater than a set threshold, the larger value between the first degree of similarity and the second degree of similarity is determined.
  • the second determining unit 54 is further configured to:
  • any one of the second abstract and the third abstract is selected as the abstract translation.
  • the device further includes:
  • the third determining unit is configured to determine the first text and determine the second text.
  • the third determining unit is further configured to:
  • the third determining unit is further configured to:
  • the third determining unit is further configured to:
  • the device further includes:
  • the display unit is configured to display the abstract translation of the first text while the speaker is speaking; wherein the content of the speaker's speech is related to the text content of the first text.
  • the generation unit 51, the translation unit 52, the first determination unit 53, the second determination unit 54, and the third determination unit may be a processor in an electronic device, such as a central processing unit.
  • CPU Central Processing Unit
  • DSP Digital Signal Processor
  • MCU Microcontroller Unit
  • FPGA Field-Programmable Gate Array
  • the display unit It can be realized through the communication interface in the electronic device.
  • the abstract generation device in the above embodiment can be correspondingly installed in the system shown in FIG. 1, that is, respectively correspondingly installed on the machine simultaneous interpretation server of the system in FIG. 1, so as to apply the abstract generation method of the embodiment of this application to In the system shown in Figure 1.
  • FIG. 6 is a schematic diagram of the hardware composition structure of an electronic device according to an embodiment of the application.
  • the electronic device 60 includes a memory 63, a processor 62, and a computer program stored on the memory 63 and running on the processor 62;
  • the processor 62 implements the method provided by one or more technical solutions when the processor 62 executes the program.
  • the processor 62 in the electronic device 60 executes the program, it realizes: generating a first abstract about the first text, and generating a second abstract about the second text; the second text is the first text Corresponding translated text; translate the first abstract to obtain a third abstract, and translate the second abstract to obtain a fourth abstract; determine the first abstract between the second abstract and the third abstract A degree of similarity, and a second degree of similarity between the first summary and the fourth summary is determined; based on the first degree of similarity and the second degree of similarity, the second summary or the first summary is selected Three abstracts are used as abstract translations of the first text; wherein, the first text, the first abstract, and the fourth abstract correspond to the first language; the second text, the second abstract, and the The third abstract corresponds to the second language.
  • the electronic device 60 further includes a communication interface 61; various components in the electronic device 60 are coupled together through the bus system 64. It can be understood that the bus system 64 is configured to implement connection and communication between these components. In addition to the data bus, the bus system 64 also includes a power bus, a control bus, and a status signal bus.
  • the memory in the embodiment of FIG. 6 above may be a volatile memory or a non-volatile memory, and may also include both volatile and non-volatile memory.
  • the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read- Only Memory, Electrically Erasable Programmable Read-Only Memory (EEPROM), Ferromagnetic Random Access Memory (FRAM), Flash Memory, Magnetic Surface Memory , CD-ROM, or CD-ROM (Compact Disc Read-Only Memory); magnetic surface memory can be magnetic disk storage or tape storage.
  • the volatile memory may be a random access memory (RAM, Random Access Memory), which is used as an external cache.
  • RAM random access memory
  • SRAM static random access memory
  • SSRAM synchronous static random access memory
  • Synchronous Static Random Access Memory Synchronous Static Random Access Memory
  • DRAM Dynamic Random Access Memory
  • SDRAM Synchronous Dynamic Random Access Memory
  • DDRSDRAM Double Data Rate Synchronous Dynamic Random Access Memory
  • ESDRAM Enhanced Synchronous Dynamic Random Access Memory
  • SLDRAM synchronous connection dynamic random access memory
  • DRRAM Direct Rambus Random Access Memory
  • the memories described in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.
  • the methods disclosed in the foregoing embodiments of the present application may be applied to a processor or implemented by a processor.
  • the processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in the form of software.
  • the aforementioned processor may be a general-purpose processor, DSP, or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the like.
  • the processor may implement or execute the methods, steps, and logical block diagrams disclosed in the embodiments of the present application.
  • the general-purpose processor may be a microprocessor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor.
  • the software module may be located in a storage medium, and the storage medium is located in a memory.
  • the processor reads the information in the memory and completes the steps of the foregoing method in combination with its hardware.
  • the embodiments of the present application also provide a storage medium, which is specifically a computer storage medium, and more specifically, a computer-readable storage medium.
  • Computer instructions that is, computer programs, are stored thereon, and the methods provided by one or more of the above technical solutions when the computer instructions are executed by the processor.
  • the disclosed method and smart device can be implemented in other ways.
  • the device embodiments described above are merely illustrative.
  • the division of the units is only a logical function division, and there may be other divisions in actual implementation, such as: multiple units or components can be combined, or It can be integrated into another system, or some features can be ignored or not implemented.
  • the coupling, or direct coupling, or communication connection between the components shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms. of.
  • the units described above as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • the functional units in the embodiments of the present application may all be integrated into a second processing unit, or each unit may be individually used as a unit, or two or more units may be integrated into one unit;
  • the above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional units.
  • the foregoing program can be stored in a computer readable storage medium. When the program is executed, it is executed. Including the steps of the foregoing method embodiment; and the foregoing storage medium includes: various media that can store program codes, such as a mobile storage device, ROM, RAM, magnetic disk, or optical disk.
  • the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it may also be stored in a computer readable storage medium.
  • the computer software product is stored in a storage medium and includes several instructions for A computer device (which may be a personal computer, a server, or a network device, etc.) executes all or part of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: removable storage devices, ROM, RAM, magnetic disks, or optical disks and other media that can store program codes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

一种摘要生成方法、装置、电子设备和存储介质。其中,摘要生成方法包括:生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要(201);所述第二文本为所述第一文本对应的翻译文本;对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要(202);确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度(203);基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文(204);其中,所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。

Description

摘要生成方法、装置、电子设备和存储介质 技术领域
本申请涉及机器翻译技术,具体涉及一种摘要生成方法、装置、电子设备和存储介质。
背景技术
在跨语种的演讲场合或会议场合中,发言者的发言时间较长,发言内容较多,因此,除了对发言内容进行同声传译,往往还会生成与发言内容相关的摘要译文,以供与会人员快速了解发言内容。相关技术中,生成的摘要译文无法准确地体现发言内容。
发明内容
为解决相关技术问题,本申请实施例提供了一种摘要生成方法、装置、电子设备和存储介质。
本申请实施例提供了一种摘要生成方法,包括:
生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;
对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;
确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;
基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,
所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
其中,上述方案中,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
确定所述第一相似度与所述第二相似度之间的较大值;
当所述较大值为所述第一相似度时,选择所述第三摘要作为所述摘要译文;
当所述较大值为所述第二相似度时,选择所述第二摘要作为所述摘要译文。
上述方案中,所述确定所述第一相似度与所述第二相似度之间的较大值,包括:
确定所述第一相似度与所述第二相似度之间的差值;
当所述差值大于设定阈值时,确定所述第一相似度与所述第二相似度之间的较大值。
上述方案中,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
确定所述第一相似度与所述第二相似度之间的差值;
当所述差值小于或等于设定阈值时,选择所述第二摘要和所述第三摘要中的任意一个作为所述摘要译文。
上述方案中,所述方法还包括:
确定所述第一文本,以及确定所述第二文本。
上述方案中,所述确定所述第一文本,包括:
对发言者的发言内容进行语音采集,得到第一语音;
对所述第一语音进行语音识别,得到所述第一文本。
上述方案中,所述确定所述第二文本,包括:
对所述第一文本进行翻译,得到所述第二文本。
上述方案中,所述确定所述第二文本,包括:
对所述发言内容对应的同声传译内容进行语音采集,得到第二语音;
对所述第二语音进行语音识别,得到所述第二文本。
上述方案中,所述方法还包括:
在发言者发言的同时显示所述第一文本的摘要译文;其中,所述发言者的发言内容与所述第一文本的文本内容相关。
本申请实施例还提供了一种摘要生成装置,所述装置包括:
生成单元,配置为生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;
翻译单元,配置为对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;
第一确定单元,配置为确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;
第二确定单元,配置为基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,
所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
本申请实施例还提供了一种电子设备,包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序,所述处理器执行所述计算机程序时,实现上述任一摘要生成方法的步骤。
本申请实施例还提供了一种存储介质,其上存储有计算机指令,所述计算机指令被处理器执行时实现上述任一摘要生成方法的步骤。
本申请实施例提供的摘要生成方法、装置、电子设备和存储介质,生成关于第一文本的第一摘要,以及生成关于第一文本对应的翻译文本的第二摘要,对第一摘要进行翻译,得到第三摘要,以及对第二摘要进行回翻,得到第四摘要,之后,基于第二摘要与第三摘要之间的第一相似度,以及第一摘要与第四摘要之间的第二相似度,对两个相似度进行比较,根据比较结果选择第二摘要或第三摘要作为第一文本的摘要译文,从而使得最终确定出的摘要译文能够更为准确地反映出第一文本的内容。在跨语种的演讲场合或会议场合中,采用本申请实施例的方案能够生成更为准确地反映发言内容的摘要,提高了信息传递的准确性。
附图说明
图1为相关技术中摘要生成方法应用的系统架构示意图;
图2为本申请实施例的摘要生成方法的一种流程示意图;
图3为本申请实施例的摘要生成方法的一种流程示意图;
图4为本申请实施例的摘要生成方法的一种流程示意图;
图5为本申请实施例的摘要生成装置的组成结构示意图;
图6为本申请实施例的电子设备的硬件组成结构示意图。
具体实施方式
在对本申请实施例的技术方案进行详细说明之前,首先对相关技术中的摘要生成方法应用的系统进行简单说明。
图1为相关技术中摘要生成方法应用的系统架构示意图。如图1所示,所述系统可包括:机器同传服务端、语音识别服务器、翻译服务器、移动端下发服务器、观众移动端、电脑(PC,Personal Computer)客户端、显示屏幕。
实际应用中,发言者可以通过PC客户端进行会议发言,并将展示的文档,如演示文稿软件(PPT,PowerPoint)的文档,投屏到所述显示屏幕,通过显示屏幕展示给用户。在进行会议发言的过程中,PC客户端采集发言者的音频,将采集的音频发送给机器同传服务端,所述机器同传服务端通过语音识别服务器对音频数据进行识别,得到识别文本,再通过翻译服务器对所述识别文本进行翻译,得到翻译结果;机器同传服务端将翻译结果发送给PC客户端,并且通过移动端下发服务器将翻译结果发送给观众移动端,为用户展示翻译结果,从而实现将发言者的发言内容翻译成用户需要的语种并进行展示。
此外,相关技术的方案中,机器同传服务端还可根据翻译服务器的翻 译结果生成关于发言内容的摘要译文,并通过移动端下发服务器将摘要译文发送给观众移动端,以帮助用户快速了解发言者的发言内容。但是,以跨语种的演讲场合或跨语种的会议场合为例,每个发言者的发言时间较长,发言内容较多,在发言内容较多时,翻译服务器对发言内容进行机器翻译,得到的翻译结果中多多少少会存在翻译错误,那么在此情况下,生成的摘要译文也就无法准确地体现发言内容,从而对用户准确理解发言内容起到负面作用,降低了信息传递的准确性。
基于此,在本申请的各种实施例中,采用两种不同的方式得到关于发言内容的两段摘要译文,并通过译文回翻、比较不同摘要原文之间的相似度及比较不同摘要译文之间的相似度的方式,最终从两段摘要译文中选择出一段作为关于发言内容的摘要译文,从而选取出更能准确地体现发言内容的摘要译文,准确地实现了信息传递。
下面结合附图及具体实施例对本申请作进一步详细的说明。
本申请实施例提供了一种摘要生成方法,应用于电子设备。图2为本申请实施例的摘要生成方法的一种流程示意图。如图2所示,所述方法包括:
步骤201:生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本。
其中,第二文本为第一文本对应的翻译文本,即第一文本为原文,第二文本为对应的译文。这里,基于第一文本和第二文本,分别生成第一摘要和第二摘要。在生成摘要时,可以采用的方式可以包括以下至少一种:
通过抽取式(extractive)的方式生成摘要;
通过生成式(abstractive)的方式生成摘要。
其中,通过抽取式的方式生成摘要,是指从文本中抽取一些具有代表性的文本片段构成摘要,这些片段可以是整个文本中的词语、句子、段落或者小节。在抽取过程中,先对文本中的片段进行排序,选择排名靠前的片段,再去除选择出的片段中的冗余信息,最终生成摘要。通过生成式的方式生成摘要,是指通过建立抽象的语意表示,使用自然语言生成技术形成摘要。对于抽取式摘要来说,与文本的相似度较高,但可读性较差,信息量较低,对于生成式摘要来说,具有较高的可读性和丰富的信息,但与文本的相似度较低,在实际应用中,也可以将抽取式的方式和生成式的方式相结合,比如先通过抽取式的方式对文本进行处理,得到抽取式摘要,再通过生成式的方式对得到的抽取式摘要进行处理,从而得到能够兼顾文本的相似度与可读性、信息量这几个指标的摘要。
在一个实施例中,所述方法还包括:
确定所述第一文本,以及确定所述第二文本。
其中,在生成第一摘要和第二摘要之前,先确定第一文本及第二文本。这里,第一文本为原文,第一文本可以为发言者的发言文稿,此外,在一个实施例中,所述确定所述第一文本,包括:
对发言者的发言内容进行语音采集,得到第一语音;
对所述第一语音进行语音识别,得到所述第一文本。
这里,在发言者进行发言时,可以通过语音采集模块,如麦克风,对发言者的声音进行采集,得到第一语音,再通过语音识别将第一语音转换为第一文本,由此得到发言者的实际发言内容对应的原文。
此外,在得到第一文本之后,还可以对第一文本进行添加标点、文本顺滑等后处理,添加标点使得第一文本能够实现合理断句,文本顺滑能够调整语音识别出的文本的词语顺序,或者为语音识别出的文本添加连词、助词等,以克服语音识别出的文本口语化的缺陷,得到符合语法表述的第一文本。
第二文本为第一文本对应的译文,在一个实施例中,所述确定所述第二文本,包括:
对所述第一文本进行翻译,得到所述第二文本。
其中,第二文本基于第一文本生成,通过对确定出的第一文本进行翻译,将得到的第一文本对应的译文作为第二文本。这里,第二文本对于发言内容的翻译质量在很大程度上与第一文本对发言内容的描述准确度相关。
在一个实施例中,所述确定所述第二文本,包括:
对所述发言内容对应的同声传译内容进行语音采集,得到第二语音;
对所述第二语音进行语音识别,得到所述第二文本。
当发言者发言时,若对发言内容进行了同声传译,那么可以通过语音采集模块,如麦克风,对同声传译的声音进行采集,得到第二语音,再通过语音识别将第二语音转换为第二文本,由此得到发言者的实际发言内容对应的译文。
这里,第二文本对于发言内容的翻译质量不再与第一文本对发言内容的描述准确度相关,而是与同声传译的翻译质量相关。
此外,在通过对同声传译内容进行语音识别的方式得到第二文本之后,也需要对第二文本进行添加标点、文本顺滑等后处理,以得到符合语法表述的第二文本。
实际应用中,可以根据应用场景选择相应的方式获取第二文本。例如,对于发言者预先提供了发言文稿的情况,可以通过对第一文本进行翻译,得到第二文本;又例如,对于发言者是即兴发言的情况,可以通过对同声传译的内容进行语音识别,进而确定出第二文本。
步骤202:对所述第一摘要进行翻译,得到第三摘要,以及对所述 第二摘要进行翻译,得到第四摘要。
这里,第一摘要为关于第一文本的摘要,为摘要原文,第二摘要为关于第二文本的摘要,为摘要译文。因此,对第一摘要进行翻译,得到的第三摘要为关于第一文本的摘要译文,对第二摘要进行翻译,得到的第四摘要为关于第二文本的摘要原文。
为了方便说明,上述第一文本、第一摘要和第四摘要均对应第一语种,也即发言者发言时所使用的语种,上述第二文本、第二摘要和第三摘要均对应第二语种,也即翻译语种。
在实际应用时,可以通过机器翻译技术进行由第一语种至第二语种的翻译,以及实现由第二语种至第一语种的回翻。具体地,可由服务器采用基于规则(Rule-Based)的机器翻译技术或者采用基于语料库(Corpus-Based)的机器翻译技术进行翻译。
步骤203:确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度。
由于第一摘要、第二摘要、第三摘要和第四摘要均为文本,因此,实际应用时,可以采用双语互译质量辅助工具(BLEU,Bilingual Evaluation Understudy)或者ROUGE(Recall-oriented Understanding for Gisting Evaluation)来计算第二摘要与第三摘要之间,或者计算第一摘要与第四摘要之间的相似度。其中,BLEU为用于衡量两个文本之间的相似程度的指标,最终得到的相似度的取值范围在0-1,取值越靠近1表示两个文本在语义上越相似;ROUGE通过将两个文本进行比较计算,得出相应的分值,以衡量两个文本之间的相似度。
步骤204:基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文。
在一个实施例中,如图3所示,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
步骤2041:确定所述第一相似度与所述第二相似度之间的较大值;
步骤2042:当所述较大值为所述第一相似度时,选择所述第三摘要作为所述摘要译文;当所述较大值为所述第二相似度时,选择所述第二摘要作为所述摘要译文。
这里,将第一相似度与第二相似度进行比较,当比较结果为第一相似度较大时,也即同为翻译语种(第二语种)的第二摘要和第三摘要之间的相似度较高时,说明生成的关于第一文本的摘要原文(第一摘要)的质量较高,且基于该摘要原文翻译得到的摘要译文(第三摘要)的翻译质量也较高,第三摘要反映第一文本内容的准确性较高,因此将第三摘要作为第一文本的摘要译文;当比较结果为第二相似度较大时,也即同为原文语种(第一语种)的第一摘要和第四摘要之间的相似度较高时, 说明基于第一文本翻译得到的第二文本的翻译质量较高,且生成的关于第二文本的第二摘要的质量也较高,第二摘要反映第一文本内容的准确性较高,因此将第二摘要作为第一文本的摘要译文。
在一个实施例中,所述确定所述第一相似度与所述第二相似度之间的较大值,包括:
确定所述第一相似度与所述第二相似度之间的差值;
当所述差值大于设定阈值时,确定所述第一相似度与所述第二相似度之间的较大值。
这里,当确定第一相似度与第二相似度之间的差值大于设定阈值时,认为第二摘要和第三摘要之间的语义相似度是不够大的,那么才进一步通过确定第一相似度与第二相似度之间的较大值,来选择将第二摘要或第三摘要作为第一文本的摘要译文。而当第一相似度与第二相似度之间的差值小于设定阈值时,认为第二摘要和第三摘要之间的语义相似度较高,那么在这种情况下,可以选择第二摘要或第三摘要中的任意一个作为第一文本的摘要译文。如图4所示,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
步骤2043:确定所述第一相似度与所述第二相似度之间的差值;
步骤2044:当所述差值小于或等于设定阈值时,选择所述第二摘要和所述第三摘要中的任意一个作为所述摘要译文。
在一个实施例中,所述方法还包括:
在发言者发言的同时显示所述第一文本的摘要译文;其中,所述发言者的发言内容与所述第一文本的文本内容相关。
这里,第一文本的文本内容即为发言者的发言内容,第一文本可以为发言者的发言文稿,也可以为通过对发言者的音频进行实时语音采集,再通过语音识别而得到的识别文本。在发言者发言的同时,生成第一文本的摘要译文,这样一来,能够在会议中更好地帮助与会人员实时地了解发言者的发言内容,把握发言要点,从而实现了信息准确、快速的传递。
本申请实施例提供的摘要生成方法,生成关于第一文本的第一摘要,以及生成关于第一文本对应的翻译文本的第二摘要,对第一摘要进行翻译,得到第三摘要,以及对第二摘要进行回翻,得到第四摘要,之后,基于第二摘要与第三摘要之间的第一相似度,以及第一摘要与第四摘要之间的第二相似度,对两个相似度进行比较,根据比较结果选择第二摘要或第三摘要作为第一文本的摘要译文,从而使得最终确定出的摘要译文能够更为准确地反映出第一文本的内容。在跨语种的演讲场合或会议场合中,采用本申请实施例的方案能够生成更为准确地反映发言内容的摘要,提高了信息传递的准确性。在同声传译的场景之下,本申请 实施例的方案能够在一定程度上避免因为机器翻译的翻译错误而带来的信息传递错误,更为准确地生成关于发言内容的摘要。
为实现本申请实施例的翻译方法,本申请实施例还提供了一种摘要生成装置,可以设置于电子设备中。图5为本申请实施例的摘要生成装置的组成结构示意图。如图5所示,所述摘要生成装置包括:
生成单元51,配置为生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;
翻译单元52,配置为对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;
第一确定单元53,配置为确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;
第二确定单元54,配置为基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,
所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
在一实施例中,所述第二确定单元54进一步配置为:
确定所述第一相似度与所述第二相似度之间的较大值;
当所述较大值为所述第一相似度时,选择所述第三摘要作为所述摘要译文;
当所述较大值为所述第二相似度时,选择所述第二摘要作为所述摘要译文。
在一实施例中,所述第二确定单元54确定所述第一相似度与所述第二相似度之间的较大值时,进一步配置为:
确定所述第一相似度与所述第二相似度之间的差值;
当所述差值大于设定阈值时,确定所述第一相似度与所述第二相似度之间的较大值。
在一实施例中,所述第二确定单元54进一步配置为:
确定所述第一相似度与所述第二相似度之间的差值;
当所述差值小于或等于设定阈值时,选择所述第二摘要和所述第三摘要中的任意一个作为所述摘要译文。
在一实施例中,所述装置还包括:
第三确定单元,配置为确定所述第一文本,以及确定所述第二文本。
在一实施例中,所述第三确定单元进一步配置为:
对发言者的发言内容进行语音采集,得到第一语音;
对所述第一语音进行语音识别,得到所述第一文本。
在一实施例中,所述第三确定单元进一步配置为:
对所述第一文本进行翻译,得到所述第二文本。
在一实施例中,所述第三确定单元进一步配置为:
对所述发言内容对应的同声传译内容进行语音采集,得到第二语音;
对所述第二语音进行语音识别,得到所述第二文本。
在一实施例中,所述装置还包括:
显示单元,配置为在发言者发言的同时显示所述第一文本的摘要译文;其中,所述发言者的发言内容与所述第一文本的文本内容相关。
实际应用时,所述生成单元51、所述翻译单元52、所述第一确定单元53、所述第二确定单元54、所述第三确定单元可由电子设备中的处理器,比如中央处理器(CPU,Central Processing Unit)、数字信号处理器(DSP,Digital Signal Processor)、微控制单元(MCU,Microcontroller Unit)或可编程门阵列(FPGA,Field-Programmable Gate Array)等实现;所述显示单元可通过电子设备中的通信接口实现。
需要说明的是:上述实施例提供的摘要生成装置在进行摘要生成时,仅以上述各程序模块的划分进行举例说明,实际应用中,可以根据需要而将上述处理分配由不同的程序模块完成,即将电子设备内部结构划分成不同的程序模块,以完成以上描述的全部或者部分处理。另外,上述实施例提供的摘要生成装置与摘要生成方法实施例属于同一构思,其具体实现过程详见方法实施例,这里不再赘述。
上文实施例中的摘要生成装置可以对应设置于图1所示的系统中,即,分别对应设置在图1系统的机器同传服务端上,以将本申请实施例的摘要生成方法应用在图1所示的系统中。
基于上述设备的硬件实现,本申请实施例还提供了一种电子设备。图6为本申请实施例的电子设备的硬件组成结构示意图,如图6所示,电子设备60包括存储器63、处理器62及存储在存储器63上并可在处理器62上运行的计算机程序;处理器62执行所述程序时实现上述一个或多个技术方案提供的方法。
具体地,电子设备60中的处理器62执行所述程序时实现:生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
需要说明的是,处理器62执行所述程序时实现的具体步骤已在上文实施例中详述,这里不再赘述。
可以理解,电子设备60还包括通信接口61;电子设备60中的各个组件通过总线系统64耦合在一起。可理解,总线系统64配置为实现这些组件之间的连接通信。总线系统64除包括数据总线之外,还包括电源总线、控制总线和状态信号总线等。
可以理解,上文图6实施例中的存储器可以是易失性存储器或非易失性存储器,也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是只读存储器(ROM,Read Only Memory)、可编程只读存储器(PROM,Programmable Read-Only Memory)、可擦除可编程只读存储器(EPROM,Erasable Programmable Read-Only Memory)、电可擦除可编程只读存储器(EEPROM,Electrically Erasable Programmable Read-Only Memory)、磁性随机存取存储器(FRAM,ferromagnetic random access memory)、快闪存储器(Flash Memory)、磁表面存储器、光盘、或只读光盘(CD-ROM,Compact Disc Read-Only Memory);磁表面存储器可以是磁盘存储器或磁带存储器。易失性存储器可以是随机存取存储器(RAM,Random Access Memory),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(SRAM,Static Random Access Memory)、同步静态随机存取存储器(SSRAM,Synchronous Static Random Access Memory)、动态随机存取存储器(DRAM,Dynamic Random Access Memory)、同步动态随机存取存储器(SDRAM,Synchronous Dynamic Random Access Memory)、双倍数据速率同步动态随机存取存储器(DDRSDRAM,Double Data Rate Synchronous Dynamic Random Access Memory)、增强型同步动态随机存取存储器(ESDRAM,Enhanced Synchronous Dynamic Random Access Memory)、同步连接动态随机存取存储器(SLDRAM,SyncLink Dynamic Random Access Memory)、直接内存总线随机存取存储器(DRRAM,Direct Rambus Random Access Memory)。本申请实施例描述的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
上述本申请实施例揭示的方法可以应用于处理器中,或者由处理器实现。处理器可能是一种集成电路芯片,具有信号的处理能力。在实现过程中,上述方法的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以是通用处理器、DSP,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。处理器可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤,可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于存储介质中,该 存储介质位于存储器,处理器读取存储器中的信息,结合其硬件完成前述方法的步骤。
本申请实施例还提供了一种存储介质,具体为计算机存储介质,更具体的为计算机可读存储介质。其上存储有计算机指令,即计算机程序,该计算机指令被处理器执行时上述一个或多个技术方案提供的方法。
在本申请所提供的几个实施例中,应该理解到,所揭露的方法和智能设备,可以通过其它的方式实现。以上所描述的设备实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,如:多个单元或组件可以结合,或可以集成到另一个系统,或一些特征可以忽略,或不执行。另外,所显示或讨论的各组成部分相互之间的耦合、或直接耦合、或通信连接可以是通过一些接口,设备或单元的间接耦合或通信连接,可以是电性的、机械的或其它形式的。
上述作为分离部件说明的单元可以是、或也可以不是物理上分开的,作为单元显示的部件可以是、或也可以不是物理单元,即可以位于一个地方,也可以分布到多个网络单元上;可以根据实际的需要选择其中的部分或全部单元来实现本实施例方案的目的。
另外,在本申请各实施例中的各功能单元可以全部集成在一个第二处理单元中,也可以是各单元分别单独作为一个单元,也可以两个或两个以上单元集成在一个单元中;上述集成的单元既可以采用硬件的形式实现,也可以采用硬件加软件功能单元的形式实现。
本领域普通技术人员可以理解:实现上述方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成,前述的程序可以存储于一计算机可读取存储介质中,该程序在执行时,执行包括上述方法实施例的步骤;而前述的存储介质包括:移动存储设备、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
或者,本申请上述集成的单元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取存储介质中。基于这样的理解,本申请实施例的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机、服务器、或者网络设备等)执行本申请各个实施例所述方法的全部或部分。而前述的存储介质包括:移动存储设备、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
需要说明的是:“第一”、“第二”等是用于区别类似的对象,而不必用于描述特定的顺序或先后次序。
另外,本申请实施例所记载的技术方案之间,在不冲突的情况下,可以任意组合。
以上所述,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,可轻易想到变化或替换,都应涵盖在本申请的保护范围之内。

Claims (12)

  1. 一种摘要生成方法,包括:
    生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;
    对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;
    确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;
    基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,
    所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
  2. 根据权利要求1所述的方法,其中,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
    确定所述第一相似度与所述第二相似度之间的较大值;
    当所述较大值为所述第一相似度时,选择所述第三摘要作为所述摘要译文;
    当所述较大值为所述第二相似度时,选择所述第二摘要作为所述摘要译文。
  3. 根据权利要求2所述的方法,其中,所述确定所述第一相似度与所述第二相似度之间的较大值,包括:
    确定所述第一相似度与所述第二相似度之间的差值;
    当所述差值大于设定阈值时,确定所述第一相似度与所述第二相似度之间的较大值。
  4. 根据权利要求1所述的方法,其中,所述基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文,包括:
    确定所述第一相似度与所述第二相似度之间的差值;
    当所述差值小于或等于设定阈值时,选择所述第二摘要和所述第三摘要中的任意一个作为所述摘要译文。
  5. 根据权利要求1所述的方法,其中,所述方法还包括:
    确定所述第一文本,以及确定所述第二文本。
  6. 根据权利要求5所述的方法,其中,所述确定所述第一文本,包括:
    对发言者的发言内容进行语音采集,得到第一语音;
    对所述第一语音进行语音识别,得到所述第一文本。
  7. 根据权利要求5所述的方法,其中,所述确定所述第二文本,包括:
    对所述第一文本进行翻译,得到所述第二文本。
  8. 根据权利要求6所述的方法,其中,所述确定所述第二文本,包括:
    对所述发言内容对应的同声传译内容进行语音采集,得到第二语音;
    对所述第二语音进行语音识别,得到所述第二文本。
  9. 根据权利要求1所述的方法,其中,所述方法还包括:
    在发言者发言的同时显示所述第一文本的摘要译文;其中,所述发言者的发言内容与所述第一文本的文本内容相关。
  10. 一种摘要生成装置,所述装置包括:
    生成单元,配置为生成关于第一文本的第一摘要,以及生成关于第二文本的第二摘要;所述第二文本为所述第一文本对应的翻译文本;
    翻译单元,配置为对所述第一摘要进行翻译,得到第三摘要,以及对所述第二摘要进行翻译,得到第四摘要;
    第一确定单元,配置为确定所述第二摘要与所述第三摘要之间的第一相似度,以及确定所述第一摘要与所述第四摘要之间的第二相似度;
    第二确定单元,配置为基于所述第一相似度与所述第二相似度,选择所述第二摘要或所述第三摘要作为所述第一文本的摘要译文;其中,
    所述第一文本、所述第一摘要和所述第四摘要对应第一语种;所述第二文本、所述第二摘要和所述第三摘要对应第二语种。
  11. 一种电子设备,包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序,所述处理器执行所述计算机程序时,实现权利要求1至9任一项所述方法的步骤。
  12. 一种存储介质,其上存储有计算机指令,所述计算机指令被处理器执行时实现权利要求1至9任一项所述方法的步骤。
PCT/CN2019/117230 2019-11-11 2019-11-11 摘要生成方法、装置、电子设备和存储介质 WO2021092730A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201980101002.0A CN114514529A (zh) 2019-11-11 2019-11-11 摘要生成方法、装置、电子设备和存储介质
PCT/CN2019/117230 WO2021092730A1 (zh) 2019-11-11 2019-11-11 摘要生成方法、装置、电子设备和存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/117230 WO2021092730A1 (zh) 2019-11-11 2019-11-11 摘要生成方法、装置、电子设备和存储介质

Publications (1)

Publication Number Publication Date
WO2021092730A1 true WO2021092730A1 (zh) 2021-05-20

Family

ID=75911538

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/117230 WO2021092730A1 (zh) 2019-11-11 2019-11-11 摘要生成方法、装置、电子设备和存储介质

Country Status (2)

Country Link
CN (1) CN114514529A (zh)
WO (1) WO2021092730A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103092829A (zh) * 2011-10-27 2013-05-08 北京百度网讯科技有限公司 一种复述资源获取方法及系统
US20140172411A1 (en) * 2012-12-17 2014-06-19 Electronics And Telecommunications Research Institute Apparatus and method for verifying context
US20140288915A1 (en) * 2013-03-19 2014-09-25 Educational Testing Service Round-Trip Translation for Automated Grammatical Error Correction
CN108804428A (zh) * 2018-06-12 2018-11-13 苏州大学 一种译文中术语错译的纠正方法、系统及相关装置
CN109829170A (zh) * 2017-11-23 2019-05-31 三星电子株式会社 机器翻译方法和设备
CN110069790A (zh) * 2019-05-10 2019-07-30 东北大学 一种通过译文回译对照原文的机器翻译系统及方法
CN110209802A (zh) * 2019-06-05 2019-09-06 北京金山数字娱乐科技有限公司 一种提取摘要文本的方法及装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103092829A (zh) * 2011-10-27 2013-05-08 北京百度网讯科技有限公司 一种复述资源获取方法及系统
US20140172411A1 (en) * 2012-12-17 2014-06-19 Electronics And Telecommunications Research Institute Apparatus and method for verifying context
US20140288915A1 (en) * 2013-03-19 2014-09-25 Educational Testing Service Round-Trip Translation for Automated Grammatical Error Correction
CN109829170A (zh) * 2017-11-23 2019-05-31 三星电子株式会社 机器翻译方法和设备
CN108804428A (zh) * 2018-06-12 2018-11-13 苏州大学 一种译文中术语错译的纠正方法、系统及相关装置
CN110069790A (zh) * 2019-05-10 2019-07-30 东北大学 一种通过译文回译对照原文的机器翻译系统及方法
CN110209802A (zh) * 2019-06-05 2019-09-06 北京金山数字娱乐科技有限公司 一种提取摘要文本的方法及装置

Also Published As

Publication number Publication date
CN114514529A (zh) 2022-05-17

Similar Documents

Publication Publication Date Title
US11328133B2 (en) Translation processing method, translation processing device, and device
US11915692B2 (en) Facilitating end-to-end communications with automated assistants in multiple languages
US11734514B1 (en) Automated translation of subject matter specific documents
JP4237001B2 (ja) 文書のコロケーション誤りを自動的に検出するシステムおよび方法
WO2021134524A1 (zh) 数据处理方法、装置、电子设备和存储介质
JP2015516587A (ja) 対話から情報を抽出するデバイス
CN109343696B (zh) 一种电子书的评论方法、装置及计算机可读存储介质
US20230223009A1 (en) Language-agnostic Multilingual Modeling Using Effective Script Normalization
Jabaian et al. Comparison and combination of lightly supervised approaches for language portability of a spoken language understanding system
US20220261545A1 (en) Systems and methods for producing a semantic representation of a document
JP2023007372A (ja) 要約生成モデルの訓練方法、装置、デバイス及び記憶媒体
CN111126084B (zh) 数据处理方法、装置、电子设备和存储介质
Cho et al. Kosp2e: Korean speech to english translation corpus
Stepanov et al. The Development of the Multilingual LUNA Corpus for Spoken Language System Porting.
WO2021087665A1 (zh) 数据处理方法、装置、服务器和存储介质
WO2021092730A1 (zh) 摘要生成方法、装置、电子设备和存储介质
Utka Language Technology Platform for Public Administration
WO2023200946A1 (en) Personalizable probabilistic models
JP2006004366A (ja) 機械翻訳システム及びそのためのコンピュータプログラム
WO2021092733A1 (zh) 字幕显示方法、装置、电子设备和存储介质
WO2021120174A1 (zh) 数据处理方法、装置、电子设备和存储介质
WO2021062757A1 (zh) 同声传译方法、装置、服务器和存储介质
Specia et al. Read, spot and translate
WO2021072649A1 (zh) 翻译方法、装置、客户端、服务器和存储介质
WO2021119908A1 (zh) 数据处理方法、装置、电子设备和存储介质

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: 19952457

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19952457

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19952457

Country of ref document: EP

Kind code of ref document: A1

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 28.10.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19952457

Country of ref document: EP

Kind code of ref document: A1