WO2022134578A1 - Procédé et appareil pour déterminer une séquence de réponses - Google Patents

Procédé et appareil pour déterminer une séquence de réponses Download PDF

Info

Publication number
WO2022134578A1
WO2022134578A1 PCT/CN2021/109383 CN2021109383W WO2022134578A1 WO 2022134578 A1 WO2022134578 A1 WO 2022134578A1 CN 2021109383 W CN2021109383 W CN 2021109383W WO 2022134578 A1 WO2022134578 A1 WO 2022134578A1
Authority
WO
WIPO (PCT)
Prior art keywords
subsequence
node
text box
sequence
endpoint
Prior art date
Application number
PCT/CN2021/109383
Other languages
English (en)
Chinese (zh)
Inventor
王德勋
徐国强
Original Assignee
深圳壹账通智能科技有限公司
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 深圳壹账通智能科技有限公司 filed Critical 深圳壹账通智能科技有限公司
Publication of WO2022134578A1 publication Critical patent/WO2022134578A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/338—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content

Definitions

  • the present application relates to the field of artificial intelligence, and in particular, to a method and apparatus for determining an answer sequence.
  • VQA Visual Question Answering
  • OCR Optical Character Recognition
  • the present application provides a method and device for determining an answer sequence, a storage medium, and a computer device, the main purpose of which is to solve the problem that the structure and typesetting of scanned documents in real scenarios are very complex, resulting in the output of question answers. Text, the problem of low accuracy.
  • a method for determining an answer sequence comprising:
  • the sequence of answers is output.
  • a storage medium wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform an operation corresponding to the method for determining an answer sequence above.
  • a computer device comprising: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface complete mutual communication through the communication bus. communication;
  • the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to perform an operation corresponding to the method for determining the answer sequence above.
  • the present application provides a method and apparatus for determining an answer sequence, a storage medium, and a computer device. Compared with the prior art, the present application can automatically delete irrelevant answers in the answer sequence, thereby improving the correct rate of the output answer sequence.
  • FIG. 1 shows a flowchart of a method for determining an answer sequence provided by an embodiment of the present application
  • FIG. 2 shows a block diagram of the composition of an apparatus for determining an answer sequence provided by an embodiment of the present application
  • FIG. 3 shows a schematic structural diagram of a computer device provided by an embodiment of the present application.
  • the embodiment of the present application provides a method for determining an answer sequence, as shown in FIG. 1 , the method includes:
  • the application environment of the present application can be the text box data obtained under the visual question answering technology.
  • the visual question answering (VQA) is a new field that needs to understand text and vision. image information, and use deep learning models to automatically parse relevant questions and answers about images, such as, what's in the image? What movement is going on? Who is playing? How many players are in the image? Who are the contestants? Is it raining? and other questions, the answer can be parsed "There are 11 players playing in the image, Booker, Lonzo, Ball, Clay, Thompson, Dirk, Nowitzki, Mark, Gasol, Kevin, Le Husband, it's raining", the data obtained by parsing is determined as the acquired text box sequence.
  • the text box sequence may be stored in the root node S 0 of the binary tree storage structure, thereby obtaining a binary tree storage structure storing the current text box sequence.
  • the clustering process refers to the method of automatically dividing a bunch of unlabeled data into several categories, which belongs to the unsupervised learning method. It is called similarity. The more similar and smaller samples are grouped into one category, and finally multiple clusters are formed, so that the samples within the same cluster are highly similar, and the differences between different clusters are high.
  • the first subsequence and the second subsequence may include two subsequences split from the initial text box, and the sum of the text box clusters included in the two subsequences is all the text boxes in the initial text box sequence.
  • the endpoint subsequence is a subsequence including both a first text box and a second text box
  • the first text box may be a start text box in the initial text box sequence
  • the second text box may be The ending text box in the initial text box.
  • a piece of answer information is output as ⁇ I, am, a, boy ⁇
  • ⁇ I ⁇ can be the starting text box
  • ⁇ boy ⁇ can be the ending text frame.
  • the reading comprehension model can directly output the sequence information of the starting text box and the ending text box, by traversing all the text boxes in the first subsequence and the text boxes in the second subsequence. All text boxes look for whether both the first text box and the second text box are included.
  • the binary tree storage structure may be back-combined by calculating the minimum horizontal distance between nodes, so as to determine whether the current node is the final result node.
  • the specific backtracking process may include: a. Taking the lowest-level node as the current node, detecting whether the minimum horizontal distance of its parent node is not greater than a preset distance threshold. b. If so, take its parent node as the current node, and perform step a, when the minimum horizontal distance is greater than the preset distance threshold, determine the text box sequence saved in the current node as the answer sequence.
  • the left child node of the binary tree storage structure is used as the storage space for the final output answer by default, and the right child node of the binary tree storage structure is used as the storage space for the discarded subsequence.
  • the left child node of the root node may be denoted as S 1
  • the right child node of the root node may be denoted as S 2 .
  • the endpoint subsequence is continued to be clustered, and it is detected whether the new two subsequences obtained by clustering include the endpoint subsequence, If so, continue to cluster the endpoint subsequences, and repeat this process until there is no endpoint subsequence.
  • backtracking and merging is performed on the binary tree storage structure to obtain an answer sequence. The specific backtracking process is the same as step 104, and will not be repeated here.
  • the answer sequence can be output, and in a practical application scenario, the answer sequence can be displayed on the display screen, so that the answer sequence can be used to solve practical problems.
  • the method before the text box sequence is acquired and the text box sequence is stored in the root node S 0 of the binary tree storage structure, the method further includes: using an optical character recognition system to detect the acquired scanned document Recognition to obtain text box clusters; arranging the text box clusters according to a preset order; using a pre-trained reading comprehension model to process the arranged text box clusters to obtain a first text box and a second text box; determining the text box cluster between the first text box and the second text box as the output text box sequence.
  • the electronic version of the document can be obtained by scanning, so that the scanned document can be detected and recognized by an optical character recognition system, and a text box cluster can be obtained. , the resulting data set of a series of text boxes.
  • the text box cluster ⁇ I, am, a, boy ⁇ each text box has a position parameter, such as the position parameter of ⁇ I ⁇ can be 1, the position parameter of ⁇ am ⁇ can be 2, according to the position
  • the parameter can arrange the text boxes in the text box cluster.
  • the first text box and the second text box can be obtained, that is, the start text box and the end text box of the answer sequence, and the start text box and the end text box can be obtained.
  • Text box clusters between text boxes can be determined as text box sequences to be output.
  • the performing clustering processing on the text box sequence to obtain the first subsequence and the second subsequence includes: performing k-means clustering on the text box sequence processing to obtain the first subsequence and the second subsequence.
  • the specific process may include: randomly extracting two text boxes in the text box sequence as the first centroid and the second centroid; respectively calculating the remaining text boxes in the text box sequence and the first centroid, the first centroid and the second centroid. Euclidean distance between two centroids; divide the text box whose Euclidean distance from the first centroid is greater than the Euclidean distance from the second centroid into the first subsequence, and divide the text box with the second centroid Text boxes whose Euclidean distance is greater than the Euclidean distance from the first centroid are divided into second subsequences.
  • the Euclidean distance also known as the Euclidean distance, is the most common distance measure. It measures the absolute distance between two points in a multi-dimensional space, that is, the true distance between two points in an m-dimensional space, or a vector the natural length. Euclidean distance in 2D and 3D space is the actual distance between two points.
  • the specific calculation formula is:
  • sqrt represents the square root
  • x i and y i represent the horizontal and vertical coordinates of the vector respectively
  • the values of i from 1 to n are traversed during the accumulation process.
  • the text box sequence is clustered into two subsequences, so that the two subsequences can be used to save the useful text box sequence and the discarded text box sequence respectively, so as to delete the text box sequence. Regardless of the answer, improve the correct rate of the output answer sequence.
  • performing the retrospective merging process on the binary tree storage structure includes: searching for the corresponding parent node S i+1 according to the left child node S 2i+1 at the bottommost level of the binary tree storage structure; calculating The minimum horizontal distance between the parent node S i+1 and the sibling node S i+2 of the parent node S i+1; determine the distance between the parent node S i+1 and the parent node S i+1 Whether the minimum horizontal distance between sibling nodes S i+2 is not greater than the preset distance threshold; if not, stop backtracking and determine the subsequence in the left child node S 2i+1 as the answer sequence; if so, continue to The upper node of the binary tree storage structure is backtracked.
  • the calculating the minimum horizontal distance between the parent node S i+1 and the sibling node S i+2 of the parent node S i+1 may specifically include: acquiring the parent node S i +1 minimum x coordinate and maximum x coordinate (A1, A2); obtain the minimum x coordinate and maximum x coordinate (B1, B2) of the sibling node S i+2 of the parent node; according to the preset minimum horizontal distance formula Calculate the minimum horizontal distance between the parent node S i+1 and the parent node's sibling node S i+2 , and the minimum horizontal distance formula includes:
  • D is the minimum horizontal distance between the parent node S i+1 and the parent node's sibling node S i+2
  • A1 is the minimum x coordinate of the parent node S i+1
  • A2 is the The maximum x coordinate of the parent node S i+1
  • B1 is the minimum x coordinate of the sibling node S i+2 of the parent node S i+1
  • B2 is the sibling node S i +2 of the parent node S i+1 the maximum x coordinate of .
  • the method before repeating the clustering processing and detection steps on the endpoint subsequences in the left child node S2, the method further includes : Perform clustering processing on the point subsequences to obtain a third subsequence and a fourth subsequence; respectively calculate the minimum horizontal distance between the third subsequence, the fourth subsequence and the endpoint subsequence in the left child node S1 ; if the minimum horizontal distance is not greater than the preset distance threshold, merge and save the corresponding third subsequence or fourth subsequence into the left child node S1.
  • the third subsequence and the fourth subsequence may be two subsequences obtained by clustering the non-endpoint subsequences in the right subnode S2.
  • the process of calculating the minimum horizontal distance between the third subsequence, the fourth subsequence and the endpoint subsequence in the left child node S1 is the same as that in step 205, and details are not repeated here.
  • the preset distance threshold can be a preset distance parameter ⁇ . In practical application scenarios, it can usually be set to 30 or 40. If the minimum horizontal distance is not greater than the preset distance threshold, the corresponding third sub The sequence or the fourth subsequence is merged and stored in the left child node S1, thereby reducing the deletion of irrelevant answers and improving the accuracy of the final answer sequence.
  • the present application provides a method for determining an answer sequence, which can obtain a sequence of text boxes, save the sequence of text boxes in the root node S 0 of a binary tree storage structure; perform clustering processing on the sequence of text boxes to obtain a first subsection sequence and second subsequence; detecting whether there is an endpoint subsequence in the first subsequence and the second subsequence, wherein the endpoint subsequence is a subsequence including both the first text box and the second text box ; If not, then carry out retrospective merge processing to described binary tree storage structure, obtain answer sequence; If yes, then save the endpoint subsequence to the left child node S1 of described root node S0 , save the non-endpoint subsequence to In the right child node S2 of the root node S0 , the clustering processing and detection steps are repeatedly performed on the endpoint subsequence in the left child node S1, until there is no endpoint subsequence, the binary tree is
  • the structure is backtracked and merged to obtain an answer sequence; the answer sequence is output. Therefore, the structure and layout of scanned documents in real scenarios are very complex, resulting in the output of question answers that easily contain irrelevant text and technical problems with a low correct rate, thereby improving the correct rate of question answers.
  • an embodiment of the present application provides a device for determining an answer sequence.
  • the device includes:
  • the obtaining unit 21 is used to obtain the text box sequence, and save the text box sequence in the root node S 0 of the binary tree storage structure;
  • the processing unit 22 is configured to perform clustering processing on the text box sequence to obtain a first subsequence and a second subsequence, and detect whether there is an endpoint subsequence in the first subsequence and the second subsequence, Wherein, the endpoint subsequence is a subsequence including both the first text box and the second text box;
  • the backtracking unit 23 is used for, if not, performing backtracking and merging processing on the binary tree storage structure to obtain an answer sequence
  • the merging unit 24 is configured to store the endpoint subsequence in the left child node S1 of the root node S0 , and store the non - endpoint subsequence in the right child node S2 of the root node S0 , if so, And repeatedly perform clustering processing and detection steps to the endpoint subsequence in the left child node S1, until there is no endpoint subsequence, carry out retrospective merging processing to the binary tree storage structure to obtain an answer sequence;
  • the output unit 25 is used for outputting the answer sequence.
  • the device also includes: a calculation unit, a judgment unit,
  • the processing unit is also specifically configured to perform clustering processing on the non-terminal subsequences in the right child node S2 to obtain a third subsequence and a fourth subsequence;
  • the computing unit is specifically further configured to respectively calculate the minimum horizontal distance between the third subsequence, the fourth subsequence and the endpoint subsequence in the left child node S1 ;
  • the judging unit is configured to combine and save the corresponding third subsequence or fourth subsequence to the left child node S1 if the minimum horizontal distance is not greater than a preset distance threshold.
  • the backtracking unit includes:
  • a search module configured to search for the corresponding parent node S i+1 according to the bottom left child node S 2i+1 of the binary tree storage structure
  • a first calculation module for calculating the minimum horizontal distance between the parent node S i+1 and the sibling node S i+2 of the parent node S i+1;
  • Judging module for judging whether the minimum horizontal distance between the parent node S i+1 and the sibling node S i+2 of the parent node S i+1 is not greater than a preset distance threshold
  • the determining module is used to stop backtracking and determine the subsequence in the left child node S 2i+1 as the answer sequence if not; if so, continue backtracking to the upper node of the binary tree storage structure.
  • the first calculation module is specifically used to obtain the minimum x-coordinate and the maximum x-coordinate (A1, A2) of the parent node S i+1 ; obtain the minimum x of the sibling node S i+2 of the parent node. coordinate and the maximum x coordinate (B1, B2); calculate the minimum horizontal distance between the parent node S i+1 and the parent node’s sibling node S i+2 according to the preset minimum horizontal distance formula, the minimum horizontal distance
  • the horizontal distance formula includes:
  • D is the minimum horizontal distance between the parent node S i+1 and the parent node's sibling node S i+2
  • A1 is the minimum x coordinate of the parent node S i+1
  • A2 is the The maximum x coordinate of the parent node S i+1
  • B1 is the minimum x coordinate of the sibling node S i+2 of the parent node S i+1
  • B2 is the sibling node S i +2 of the parent node S i+1 the maximum x coordinate of .
  • processing unit is specifically configured to perform k-means clustering processing on the text box sequence to obtain a first subsequence and a second subsequence.
  • processing unit includes;
  • an extraction module for randomly extracting 2 text boxes in the text box sequence as the first centroid and the second centroid;
  • a second calculation module configured to calculate the Euclidean distances between the remaining text boxes in the text box sequence and the first and second centroids
  • a division module configured to divide a text box whose Euclidean distance from the first centroid is greater than the Euclidean distance from the second centroid into a first subsequence, and divides the Euclidean distance from the second centroid Text boxes with distances greater than the Euclidean distance from the first centroid are divided into second subsequences.
  • the device also includes:
  • the recognition unit is used for detecting and recognizing the acquired scanned document by using the optical character recognition system to obtain the text box cluster;
  • an arranging unit for arranging the text box clusters in a preset order
  • a training unit for processing the arranged text box clusters by using a pre-trained reading comprehension model to obtain a first text box and a second text box;
  • a determining unit configured to determine a text box cluster between the first text box and the second text box as an output text box sequence.
  • a storage medium may be non-volatile or volatile, and the storage medium stores at least one executable instruction, and the computer-executable instruction can execute the above-mentioned A method of determining the sequence of answers in any method embodiment.
  • FIG. 3 shows a schematic structural diagram of a computer device according to an embodiment of the present application.
  • the specific embodiment of the present application does not limit the specific implementation of the computer device.
  • the computer device may include: a processor (processor) 302, a communication interface (Communications) Interface) 304, memory 306, and communication bus 308.
  • processor processor
  • Communication interface Communication interface
  • the processor 302 , the communication interface 304 , and the memory 306 communicate with each other through the communication bus 308 .
  • the communication interface 303 is used to communicate with network elements of other devices such as clients or other servers.
  • the processor 302 is configured to execute the program 310, and specifically may execute the relevant steps in the above-mentioned embodiment of the method for determining the answer sequence.
  • the program 310 may include program code including computer operation instructions.
  • the processor 302 may be a central processing unit (CPU), or a specific integrated circuit (ASIC) (Application Specific Integrated Circuit). Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present application.
  • the one or more processors included in the computer equipment may be the same type of processors, such as one or more CPUs; or may be different types of processors, such as one or more CPUs and one or more ASICs.
  • the memory 306 is used to store the program 310 .
  • Memory 306 may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory). memory), such as at least one disk storage.
  • the program 310 can specifically be used to cause the processor 302 to perform the following operations:
  • the sequence of answers is output.
  • the text box sequence can be obtained by obtaining the text box sequence, and the text box sequence can be saved in the root node S 0 of the binary tree storage structure; the text box sequence can be clustered to obtain the first subsequence and the first subsequence Two subsequences; detect whether there is an endpoint subsequence in the first subsequence and the second subsequence, wherein the endpoint subsequence is a subsequence that includes both the first text box and the second text box; if no , then perform retrospective merge processing on the binary tree storage structure to obtain an answer sequence; if so, save the endpoint subsequence to the left child node S1 of the root node S0 , and save the non-endpoint subsequence to the root In the right child node S2 of the node S0 , the clustering processing and the detection steps are repeatedly performed on the endpoint subsequence in the left child node S1, until there is
  • modules or steps of the present application can be implemented by a general-purpose computing device, and they can be centralized on a single computing device or distributed in a network composed of multiple computing devices Alternatively, they may be implemented in program code executable by a computing device, such that they may be stored in a storage device and executed by the computing device, and in some cases, in a different order than here
  • the steps shown or described are performed either by fabricating them separately into individual integrated circuit modules, or by fabricating multiple modules or steps of them into a single integrated circuit module.
  • the present application is not limited to any particular combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Library & Information Science (AREA)
  • Mathematical Physics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

L'invention concerne un procédé et un appareil pour déterminer une séquence de réponses, ayant trait au domaine technique de l'intelligence artificielle, ayant pour principal objectif de résoudre un problème de texte non pertinent contenu dans une séquence de zones de texte de réponse, entraînant un faible taux de réponses correctes. Le procédé comprend les étapes consistant à : obtenir une séquence de zones de texte, et sauvegarder la séquence de zones de texte dans un nœud racine S0 d'une structure de stockage d'arbre binaire (S101); segmenter la séquence de zones de texte pour obtenir une première sous-séquence et une seconde sous-séquence, et détecter si des sous-séquences à point final sont présentes dans la première sous-séquence et la seconde sous-séquence (S102); si tel n'est pas le cas, alors effectuer un retour arrière sur la structure de stockage d'arbre binaire pour obtenir une séquence de réponses (S103); si tel est le cas, alors sauvegarder une sous-séquence à point final sur un nœud enfant gauche S1 du nœud racine S0, et sauvegarder une sous-séquence sans point final sur un nœud enfant droit S2 du nœud racine S0, et répéter la segmentation et la détection de sous-séquences à point final dans le nœud enfant gauche S1 jusqu'à ce qu'il n'y ait plus de sous-séquence à point final, et effectuer un retour arrière sur la structure de stockage d'arbre binaire pour obtenir et délivrer en sortie une séquence de réponses.
PCT/CN2021/109383 2020-12-22 2021-07-29 Procédé et appareil pour déterminer une séquence de réponses WO2022134578A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011529776.4A CN112667771A (zh) 2020-12-22 2020-12-22 答案序列的确定方法及装置
CN202011529776.4 2020-12-22

Publications (1)

Publication Number Publication Date
WO2022134578A1 true WO2022134578A1 (fr) 2022-06-30

Family

ID=75407700

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/109383 WO2022134578A1 (fr) 2020-12-22 2021-07-29 Procédé et appareil pour déterminer une séquence de réponses

Country Status (2)

Country Link
CN (1) CN112667771A (fr)
WO (1) WO2022134578A1 (fr)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112667771A (zh) * 2020-12-22 2021-04-16 深圳壹账通智能科技有限公司 答案序列的确定方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015062482A1 (fr) * 2013-11-01 2015-05-07 Tencent Technology (Shenzhen) Company Limited Système et procédé de réponse automatique à des questions
US20150286516A1 (en) * 2014-01-27 2015-10-08 International Business Machines Corporation Detecting an abnormal subsequence in a data sequence
CN109508733A (zh) * 2018-10-23 2019-03-22 北京邮电大学 一种基于分布概率相似度度量的异常检测方法
CN110287282A (zh) * 2019-05-20 2019-09-27 湖南大学 基于树核计算的智能对话系统应答方法及智能对话系统
US20200322694A1 (en) * 2019-04-08 2020-10-08 Spirent Communications, Inc. Training an encrypted video stream network scoring system with non-reference video scores
CN111814843A (zh) * 2020-03-23 2020-10-23 同济大学 视觉问答系统中图像特征模块端到端训练方法及应用
CN112667771A (zh) * 2020-12-22 2021-04-16 深圳壹账通智能科技有限公司 答案序列的确定方法及装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015062482A1 (fr) * 2013-11-01 2015-05-07 Tencent Technology (Shenzhen) Company Limited Système et procédé de réponse automatique à des questions
US20150286516A1 (en) * 2014-01-27 2015-10-08 International Business Machines Corporation Detecting an abnormal subsequence in a data sequence
CN109508733A (zh) * 2018-10-23 2019-03-22 北京邮电大学 一种基于分布概率相似度度量的异常检测方法
US20200322694A1 (en) * 2019-04-08 2020-10-08 Spirent Communications, Inc. Training an encrypted video stream network scoring system with non-reference video scores
CN110287282A (zh) * 2019-05-20 2019-09-27 湖南大学 基于树核计算的智能对话系统应答方法及智能对话系统
CN111814843A (zh) * 2020-03-23 2020-10-23 同济大学 视觉问答系统中图像特征模块端到端训练方法及应用
CN112667771A (zh) * 2020-12-22 2021-04-16 深圳壹账通智能科技有限公司 答案序列的确定方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LIU LU: "Research on Answer Generation Based on Sequence-to-sequence Model", MASTER THESIS, TIANJIN POLYTECHNIC UNIVERSITY, CN, no. 9, 15 September 2019 (2019-09-15), CN , XP055946865, ISSN: 1674-0246 *

Also Published As

Publication number Publication date
CN112667771A (zh) 2021-04-16

Similar Documents

Publication Publication Date Title
Mahdavi et al. ICDAR 2019 CROHME+ TFD: Competition on recognition of handwritten mathematical expressions and typeset formula detection
WO2021135910A1 (fr) Procédé d'extraction d'informations basé sur la compréhension de lecture de machine et dispositif associé
CN111259653B (zh) 基于实体关系消歧的知识图谱问答方法、系统以及终端
US10915577B2 (en) Constructing enterprise-specific knowledge graphs
US11275906B2 (en) Natural language text conversion and method therefor
CN109086654B (zh) 手写模型训练方法、文本识别方法、装置、设备及介质
CN107729865A (zh) 一种手写体数学公式离线识别方法及系统
KR102124466B1 (ko) 웹툰 제작을 위한 콘티를 생성하는 장치 및 방법
WO2018196718A1 (fr) Procédé et dispositif de désambiguïsation d'image, support de stockage et dispositif électronique
CN111460170B (zh) 一种词语识别方法、装置、终端设备及存储介质
CN111666350B (zh) 一种基于bert模型的医疗文本关系抽取的方法
CN113127605B (zh) 一种目标识别模型的建立方法、系统、电子设备及介质
CN112115232A (zh) 一种数据纠错方法、装置及服务器
WO2022206094A1 (fr) Procédé et appareil de génération d'un dispositif de sous-titrage, et procédé et appareil de production de sous-titre
WO2022134578A1 (fr) Procédé et appareil pour déterminer une séquence de réponses
JP5343617B2 (ja) 文字認識プログラム、文字認識方法および文字認識装置
WO2013097072A1 (fr) Procédé et appareil de reconnaissance d'un caractère d'une vidéo
CN117496521A (zh) 一种表格关键信息抽取方法、系统、装置及可读存储介质
CN113505787A (zh) 题目批改方法及系统、采用的电子设备及计算机可读介质
CN112818693A (zh) 一种电子元器件型号词的自动提取方法及系统
CN116189214A (zh) 版面分析方法、装置、电子设备及存储介质
CN113553415B (zh) 问答匹配的方法、装置及电子设备
CN113920291A (zh) 基于图片识别结果的纠错方法、装置、电子设备及介质
US7240050B2 (en) Methods and apparatus for generating automated graphics using stored graphics examples
CN115050025A (zh) 基于公式识别的知识点抽取方法及装置

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: 21908583

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 31.10.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21908583

Country of ref document: EP

Kind code of ref document: A1