WO2020098669A1 - 一种表情输入的方法、装置、设备以及存储介质 - Google Patents

一种表情输入的方法、装置、设备以及存储介质 Download PDF

Info

Publication number
WO2020098669A1
WO2020098669A1 PCT/CN2019/117846 CN2019117846W WO2020098669A1 WO 2020098669 A1 WO2020098669 A1 WO 2020098669A1 CN 2019117846 W CN2019117846 W CN 2019117846W WO 2020098669 A1 WO2020098669 A1 WO 2020098669A1
Authority
WO
WIPO (PCT)
Prior art keywords
expression
user
information
feature vector
session
Prior art date
Application number
PCT/CN2019/117846
Other languages
English (en)
French (fr)
Inventor
陈秋益
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2020098669A1 publication Critical patent/WO2020098669A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/02—Input arrangements using manually operated switches, e.g. using keyboards or dials
    • G06F3/023—Arrangements for converting discrete items of information into a coded form, e.g. arrangements for interpreting keyboard generated codes as alphanumeric codes, operand codes or instruction codes
    • G06F3/0233—Character input methods
    • G06F3/0237—Character input methods using prediction or retrieval techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/02—Input arrangements using manually operated switches, e.g. using keyboards or dials
    • G06F3/023—Arrangements for converting discrete items of information into a coded form, e.g. arrangements for interpreting keyboard generated codes as alphanumeric codes, operand codes or instruction codes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
    • G06F2203/01—Indexing scheme relating to G06F3/01
    • G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns

Definitions

  • the present disclosure relates to the technical field of mobile terminals, and in particular, to a method, device, device, and storage medium for expression input.
  • the technical problem solved by the solution provided by the embodiment of the present disclosure is that it is not possible to find and input a suitable expression simply and effectively during a user session.
  • An expression input method applied to a terminal, includes: during a user session, acquiring session situation information during the user session; and determining from the expression estimation model based on the session situation information A recommended expression matching the conversation situation information; displaying the recommended expression so that the user can select the recommended expression and send it to the user session.
  • An apparatus for expression input includes: an acquisition module for acquiring session scenario information during the user session during a user session; In the expression estimation model, a recommended expression matching the conversation situation information is determined; a display module is used to display the recommended expression so that the user can select the recommended expression and send it to the user session.
  • the device includes: a processor, and a memory coupled to the processor; the memory stores an expression input that can run on the processor.
  • the program for emoji input when executed by the processor, implements the steps of the method for emoji input according to an embodiment of the present disclosure.
  • a computer storage medium stores a program for expression input, and when the program for expression input is executed by a processor, the method for expression input according to an embodiment of the present disclosure is implemented A step of.
  • FIG. 1 is a flowchart of a method for expression input provided by an embodiment of the present disclosure
  • FIG. 2 is a schematic diagram of an expression input device provided by an embodiment of the present disclosure
  • FIG. 3 is a structural block diagram of an expression input system provided by an embodiment of the present disclosure.
  • FIG. 4 is a flowchart of a method for emoji recommendation provided by an embodiment of the present disclosure
  • FIG. 5 is a flowchart of a method for model learning provided by an embodiment of the present disclosure.
  • FIG. 1 is a flowchart of a method for expression input provided by an embodiment of the present disclosure. As shown in FIG. 1, it is applied to a terminal, and includes: Step S101: During a user session, obtain session scenario information during the user session; Step S102: According to the conversation scenario information, determine a recommended expression matching the conversation scenario information from the expression estimation model; Step S103: Display the recommended expression so that the user can select the recommended expression and send it to the user In conversation.
  • the display method for displaying the recommended emoticons includes any one of the following: it may be that the confirmed recommended emoticons are displayed in a pop-up window, or may be displayed in a dialog box.
  • the recommended emoticons include at least one, and when there are multiple recommended emoticons, the matching degree matching the conversation scenario information is arranged and displayed in order from high to low.
  • the session context information includes at least any of the following: user session context information, current environment information, and user physiological information, etc .; the session context information refers to the user and friends or users in the chat process of the communication social group Any or part or all of the session context information, current environment information, and user physiological information.
  • the user's conversation context information may include one or more of the following information: conversation text information, conversation scene information, conversation object information, and sent expression information; conversation text information refers to the user's chat with friends or social groups in communication Information, such as discussing birthday parties or travel arrangements, etc .; conversation scene information refers to the scene of the conversation, such as a group chat with a friend or many friends; conversation object information refers to whether the conversation partner is a conversation object or Multiple conversation objects and the degree of close friends with the conversation objects, such as friend relationship or non-friend relationship, etc .; sending expression information refers to the expression records that have been sent to the conversation object.
  • the expression records contain at least the following fields : Emoticon ID, sending object and scene, and sending time.
  • the current environmental information may include one or more of the following information: geographic location, date, weather, temperature, and application usage of the terminal; the current geographic location, date, and weather may be obtained through other applications of the terminal or connected to external devices As well as temperature, etc., the application usage of the terminal can be obtained by detecting the startup application of the terminal, such as using games, videos, music, shopping, etc.
  • the user's physiological information may include one or more of the following: user's facial expression, user's heartbeat, user's blood pressure, user's temperature, etc .; the user's facial expression can be collected through the terminal camera, and can be detected by connecting an optional external wearable device User heartbeat, user blood pressure, user temperature, etc.
  • the determining the recommended expression matching the conversation context information from the expression estimation model based on the conversation context information includes: separately calculating a user conversation context information feature vector in the conversation context information S, the current environment information feature vector H and the user physiological information feature vector L; according to the user session context information feature vector S, the current environment information feature vector H and the user physiological information feature vector L, the expression recommendation vector is calculated T; according to the expression recommendation vector T, determine the recommended expression matching the expression recommendation vector T from the expression estimation model.
  • an expression recommendation model including the conversation situation information and the user-selected expression is established, and the expression recommendation model includes conversation Situation information table and recommended expression table.
  • the conversation scenario information table includes: user session context information, current environment information and user physiological information;
  • the recommended expression table includes a variety of expression packages, expression packages include dynamic expression packages and static expression packages; expression packages can be created by themselves, or Obtained from the network side.
  • the method may further include: an operation step of updating the expression estimation model, which includes: saving the conversation situation information and the recommended expression or non-recommended expression selected by the user after determining the recommended expression Into the historical database; when it is detected that the expression estimation model needs to be updated, the historical database is used to train the expression estimation model to generate a new expression estimation model; the new expression estimation model is saved, and at the same time Delete the expression estimation model.
  • an operation step of updating the expression estimation model which includes: saving the conversation situation information and the recommended expression or non-recommended expression selected by the user after determining the recommended expression Into the historical database; when it is detected that the expression estimation model needs to be updated, the historical database is used to train the expression estimation model to generate a new expression estimation model; the new expression estimation model is saved, and at the same time Delete the expression estimation model.
  • FIG. 2 is a schematic diagram of an expression input device provided by an embodiment of the present disclosure. As shown in FIG. 2, it includes: an acquisition module 201, a speculation module 202, and a display module 203.
  • the obtaining module 201 is used to obtain the session scenario information during the user session during the user session; the speculation module 202 is used to determine the conversation scenario from the expression estimation model based on the session scenario information A recommended emoticon whose information matches; the display module 203 is configured to display the recommended emoticon so that the user can select the recommended emoticon and send it to the user session.
  • the session context information includes at least any of the following: user session context information, current environment information, and user physiological information, etc .; the user session context information includes at least any of the following: session text information, session scene information , Conversation object information, and expression information, etc .; the current environmental information includes at least any of the following: geographic location, date, weather, temperature, and application usage of the terminal; the user physiological information includes at least any of the following: the user ’s face Facial expressions, user heartbeat, user blood pressure, user temperature, etc.
  • the speculation module 203 includes: a first calculation unit for separately calculating the user session context information feature vector S, the current environment information feature vector H, and the user physiological information feature vector in the session context information L; a second calculation unit for calculating the expression recommendation vector T based on the user session context information feature vector S, the current environment information feature vector H and the user physiological information feature vector L; the speculation unit for According to the expression recommendation vector T, a recommended expression matching the expression recommendation vector T is determined from the expression estimation model.
  • the device includes: a processor, and a memory coupled to the processor; the memory stores an expression input that can run on the processor.
  • the program for emoji input when executed by the processor, implements the steps of the method for emoji input according to an embodiment of the present disclosure.
  • a computer storage medium stores a program for expression input, and when the program for expression input is executed by a processor, the method for expression input according to an embodiment of the present disclosure is implemented A step of.
  • the embodiments of the present disclosure rely on an intelligent terminal device and instant messaging software, and the intelligent terminal device is powered by a battery.
  • the intelligent terminal can support power detection, information storage, geographic location detection, has a communication module, a camera module, and can be connected to an external wearable device (optional).
  • the embodiments of the present disclosure are based on the following aspects: 1. User session context information (text information, contact information, location group information, etc.) 2. Current user physiological information (facial expression, heartbeat, blood pressure, etc.) 3. Current Environmental information (geographic location, weather, date, temperature, application usage records, etc.) 4. User history sent expression records. The diversification of input information is helpful to recommend emoticons to users more accurately.
  • the embodiment of the present disclosure also includes a self-learning function, which can comprehensively send expression records (including the current and all historical data described above) of the user, continuously improve and optimize the model, and more accurately match the individual needs of individual users.
  • FIG. 3 is a structural block diagram of an expression input system provided by an embodiment of the present disclosure. As shown in FIG. 3, it includes session context detection, environment information detection, physiological information detection, speculative model, learning model, and historical data records.
  • Session context detection and recording of instant messaging software Obtain the user ’s
  • Conversation information refers to the chat information between the user and friends or social groups in communication
  • Conversational scenes and conversational objects refer to the scenes and associated friend relationships of each conversation.
  • the scene includes the composition of the group members and the relationship between each member and the user (whether it is a friend relationship, separate from the user) Frequent conversations, etc.), assuming that the relationship between the user and group friend n is Gn, and the value of Gn is 0 to 1, if it is not a friend relationship with n, then the value of Gn is 0; if it is a friend relationship, Gn
  • the value depends on the frequency of individual chats and the frequency of interaction in the circle of friends. The more frequent the value, the closer to 1;
  • Expression sending record refers to the record of the user sending the expression in any conversation.
  • the record contains at least the following fields: expression ID, sending object and scene, and sending time.
  • the vector of L will eventually combine with other features to calculate the expression recommendation vector T.
  • speculative model based on the information provided by the session context detection, environmental information detection, physiological information detection through internal processing to calculate the expression recommendation vector T, and then more vector T associated with the user may need to send multiple expressions, and according to priority Sort.
  • the calculation expression of T is as follows:
  • W1 is the weight of the embedding recommendation session context information feature vector
  • W2 is the weight of the emoji recommendation environment information feature vector
  • W3 is the weight of the emoji recommendation environment information feature vector
  • W1 + W2 + W3 1.
  • Emoticon recommendation present the recommended emoticons to the user for selection according to priority
  • the training model is learned based on historical data records to learn the user's choice of expression sent under multiple data conditions. After the training is completed, a new speculation model is generated to replace the original speculation model.
  • the learning model is mainly used to learn each weight value W.
  • the recommendation model will use a common set of weight values for initial calculation. If the recommended expression user is not used, the user sends a non-recommended Expression, learning model will learn the behavior of the user, accumulate to a certain period of time to re-learn the model based on the user's behavior and train the new model to update the W value to replace the old recommended model, and gradually optimize in continuous learning and updating Closer to user's behavior, improve recommendation accuracy
  • FIG. 4 is a flowchart of a method for emoticon recommendation provided by an embodiment of the present disclosure. As shown in FIG. 4, it includes: S401: acquiring contextual information, environment information, and physiological information of a timely communication software session through capabilities provided by the device; S402: speculation model based on The above information speculates the emoticon that the user needs to send; S403: The emoticon recommendation interface presents the recommended emoji for the user to choose.
  • FIG. 5 is a flowchart of a method of model learning provided by an embodiment of the present disclosure. As shown in FIG. 5, it includes: S501: recording session context information, environment information, and physiological information; S502: learning a model to generate new guesses based on historical data recording training Model; S503: The new speculation model replaces the original speculation model.
  • the emoticon model can automatically recommend the emoticon to the user for the user to choose, eliminating the user's cumbersome expression search step, and through the self Learning models can continuously improve and optimize speculative models in order to better recommend emoticons to users and improve the user experience.

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

一种表情输入的方法、装置、设备以及存储介质，涉及移动终端技术领域，其方法包括：在用户会话期间，获取所述用户会话期间的会话情景信息（S101）；根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情（S102）；显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中（S103）。

Description

一种表情输入的方法、装置、设备以及存储介质
本公开要求享有2018年11月15日提交的名称为“一种表情输入的方法、装置、设备以及存储介质”的中国专利申请CN201811356978.6的优先权，其全部内容通过引用并入本文中。
技术领域
本公开涉及移动终端技术领域，特别涉及一种表情输入的方法、装置、设备以及存储介质。
背景技术
现阶段，由于通讯技术的不断发展及智能终端的大量使用，即时通讯软件已经非常普及，成为人们日常沟通交流的基本通讯工具，及时通讯软件的功能及交互方式也在不断丰富提升。沟通交互方式除了文字还包括图片、语音、视频等，其中一种特殊的沟通表达方式称之为表情，表情是人们在通过即时通讯软件交流时，用图片来代替语言文字及语音，不仅可以增加用户在沟通过程中的乐趣，还能增强表达效果。
在一些情况中，用户通过及时通讯软件发送表情时，都需要在特定菜单中人工查找合适的表情，现有操作繁琐费时，影响了用户体验。因此现有表情的发送方式需要提高。
发明内容
根据本公开实施例提供的方案解决的技术问题是在用户会话期间不能简单有效地查找并输入合适的表情。
根据本公开实施例提供的一种表情输入的方法，应用于终端上，包括：在用户会话期间，获取所述用户会话期间的会话情景信息；根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；显示所述推荐表 情，以便用户选取所述推荐表情后发送到所述用户会话中。
根据本公开实施例提供的一种表情输入的装置，包括：获取模块，用于在用户会话期间，获取所述用户会话期间的会话情景信息；推测模块，用于根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；显示模块，用于显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中。
根据本公开实施例提供的一种表情输入的设备，所述设备包括：处理器，以及与所述处理器耦接的存储器；所述存储器上存储有可在所述处理器上运行的表情输入的程序，所述表情输入的程序被所述处理器执行时实现根据本公开实施例提供的所述的表情输入的方法的步骤。
根据本公开实施例提供的一种计算机存储介质，所述存储介质存储有表情输入的程序，所述表情输入的程序被处理器执行时实现根据本公开实施例提供的所述的表情输入的方法的步骤。
附图说明
此处所说明的附图用来提供对本公开的进一步理解，构成本公开的一部分，本公开的示意性实施例及其说明用于理解本公开，并不构成对本公开的不当限定。在附图中：
图1是本公开实施例提供的一种表情输入的方法流程图；
图2是本公开实施例提供的一种表情输入的装置示意图；
图3是本公开实施例提供的表情输入的系统结构框图；
图4是本公开实施例提供的表情推荐的方法流程图；
图5是本公开实施例提供的模型学习的方法流程图。
具体实施方式
以下结合附图对本公开的优选实施例进行详细说明，应当理解，以下所说明的优选实施例仅用于说明和解释本公开，并不用于限定本公开。
图1是本公开实施例提供的一种表情输入的方法流程图，如图1所示，应用 于终端上，包括：步骤S101：在用户会话期间，获取所述用户会话期间的会话情景信息；步骤S102：根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；步骤S103：显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中。
在一实施方式中，显示所述推荐表情的显示方式包括以下任一：可以是弹出窗口显示所述确认的推荐表情，或者可以是在对话框中显示所述确认的推荐表情等。其中，推荐表情至少包括一个，当为多个推荐表情时，按照会话情景信息相匹配的匹配度从高到低依次排列显示出来。
在一实施方式中，所述会话情景信息至少包括以下任一：用户会话上下文信息、当前环境信息以及用户生理信息等；会话情景信息是指用户与好友或者在通讯社交群的聊天过程中的用户会话上下文信息、当前环境信息以及用户生理信息等中的任一或部分或所有的信息。
所述用户会话上下文信息，可以包括下述一种或多种信息：会话文本信息、会话场景信息、会话对象信息以及发送表情信息等；会话文本信息是指用户与好友或者在通讯社交群的聊天信息，比如讨论生日聚会或是旅游安排事宜等等；会话场景信息是指会话所处的场景，比如是与一个好友还是众多好友组成的群聊；会话对象信息是指会话对方是一个会话对象还是多个会话对象以及与会话对象之间的密切好友程度，比如是好友关系还是非好友关系等等；发送表情信息是指与会话对象之间已发送过的表情记录，该表情记录至少包含以下字段：表情ID、发送对象及场景以及发送时间。
所述当前环境信息可以包括下述一种或多种信息：地理位置、日期、天气、温度以及终端的应用使用情况等；可以通过终端其他应用或连接外部设备获取当前的地理位置、日期、天气以及温度等等，通过检测终端的启动应用得到终端的应用使用情况，比如正在使用游戏、视频、音乐、购物等等。
所述用户生理信息可以包括下述一种或多种：用户脸部表情、用户心跳、用户血压以及用户温度等；可以通过终端摄像头采集用户脸部表情，可以通过连接可选外部可穿戴设备检测用户心跳、用户血压以及用户温度等。
在一实施方式中，所述根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情包括：分别计算出所述会话情景信息中的用户会话上下文信息特征向量S、当前环境信息特征向量H以及用户生理信息特征向 量L；根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T；根据所述表情推荐向量T，从表情推测模型中确定与所述表情推荐向量T相匹配的推荐表情。
在一实施方式中，所述根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T包括：T＝S*W1+H*W2+L*W3；其中，W1是所述用户会话上下文信息特征向量S的权重；所述W2是所述当前环境信息特征向量H的权重；所述W3是所述用户生理信息特征向量L的权重；且W1+W2+W3＝1。
在一实施方式中，通过不断的记录众多用户的会话情景信息以及在所述会话情景下所确定的表情，建立包含会话情景信息和用户所选表情的表情推荐模型，所述表情推荐模型包括会话情景信息表和推荐表情表。其中，会话情景信息表包括：用户会话上下文信息、当前环境信息以及用户生理信息；推荐表情表中包括多种表情包，表情包包括动态表情包和静态表情包；表情包可以自己创建，也可以从网络侧获取。
在一实施方式中，该方法还可以包括：对所述表情推测模型进行更新的操作步骤，其包括：将所述会话情景信息及在确定推荐表情后用户所选取的推荐表情或非推荐表情保存到历史数据库中；当检测到需要对表情推测模型进行更新时，利用所述历史数据库对所述表情推测模型进行训练，生成新的表情推测模型；将所述新的表情推测模型进行保存，同时删除所述表情推测模型。
图2是本公开实施例提供的一种表情输入的装置示意图，如图2所示，包括：获取模块201、推测模块202以及显示模块203。
所述获取模块201，用于在用户会话期间，获取所述用户会话期间的会话情景信息；所述推测模块202，用于根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；所述显示模块203，用于显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中。
在一实施方式中，所述会话情景信息至少包括以下任一：用户会话上下文信息、当前环境信息以及用户生理信息等；所述用户会话上下文信息至少包括以下任一：会话文本信息、会话场景信息、会话对象信息以及发送表情信息等；所述当前环境信息至少包括以下任一：地理位置、日期、天气、温度以及终端的应用使用情况等；所述用户生理信息至少包括以下任一：用户脸部表情、用户心跳、 用户血压以及用户温度等。在一实施方式中，所述推测模块203包括：第一计算单元，用于分别计算出所述会话情景信息中的用户会话上下文信息特征向量S、当前环境信息特征向量H以及用户生理信息特征向量L；第二计算单元，用于根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T；推测单元，用于根据所述表情推荐向量T，从表情推测模型中确定与所述表情推荐向量T相匹配的推荐表情。
根据本公开实施例提供的一种表情输入的设备，所述设备包括：处理器，以及与所述处理器耦接的存储器；所述存储器上存储有可在所述处理器上运行的表情输入的程序，所述表情输入的程序被所述处理器执行时实现根据本公开实施例提供的所述的表情输入的方法的步骤。
根据本公开实施例提供的一种计算机存储介质，所述存储介质存储有表情输入的程序，所述表情输入的程序被处理器执行时实现根据本公开实施例提供的所述的表情输入的方法的步骤。
本公开实施例依赖于智能终端设备及即时通讯软件，该智能终端设备采用电池供电。该智能终端能够支持电量检测、信息保存、地理位置检测、具备通讯模块、拍照模块、可连接外部可穿戴设备(可选)。
本公开实施例根据以下几个方面：1.用户会话上下文信息(文本信息、联系人信息、所处会话群组信息等)2.当前用户生理信息(面部表情、心跳、血压等)3.当前环境信息(地理位置、天气、日期、温度、应用使用记录等)4.用户历史发送表情记录。输入信息的多样化有利于更加准确地向用户推荐表情。本公开实施例还包含自学习功能，可以综合用户发送表情记录(包含以上所述当前及全部历史数据)，不断改进和优化模型，更加精准匹配单个用户个性化需求。
图3是本公开实施例提供的表情输入的系统结构框图，如图3所示，包括会话上下文检测、环境信息检测、生理信息检测、推测模型、学习模型以及历史数据记录。
一、即时通讯软件会话上下文检测及记录：获取用户在及时通讯应用中的
1.会话信息：是指用户与好友或者在通讯社交群的聊天信息；
2.会话场景及会话对象：是指每一个会话的所处场景及关联好友关系，如在某群里面其场景包含群成员组成、每个成员与用户的关系(是否是好友关系，与 用户单独会话的频繁情况等)，假设用户与群好友n的关系值为Gn，Gn的取值为0到1,如果与n并非好友关系，那么Gn的取值为0；如果为好友关系，Gn的取值取决于单独聊天频次及朋友圈的交互频次，越频繁取值越趋近于1；
3.表情发送记录：是指用户在任何会话中发送表情的记录，该记录至少包含以下字段：表情ID、发送对象及场景以及发送时间。
以一个实施过程为例说明如何采用会话上下文信息进行表情推荐：在某好友群A，通过模型分析在A群中用户及其他好友的聊天信息，可以推测用户的情感倾向特征向量S1和群的整体情感氛围特征向量S2，用户与群里面的各成员的关系向量S3(G的归一化向量，可采用总关系向量的均值)，用户在此群发送表情记录的情感倾向归一化向量S4。根据以上四个特征向量可以计算表情推荐会话上下文信息特征向量S，S＝S1*Ws1+S2*Ws2+S3*Ws3+S4*Ws4，其中Wsn为各特征向量的权重。S的向量最终会结合以下其他特征计算出表情推荐向量T。
二、环境信息检测：获取地理位置H1、日期(包含节假日信息)H2、天气H3、温度H3(可选)、终端应用使用情况(是否在使用游戏H4、视频H5、音乐H6、购物H7等应用)等；根据以上特征向量可以计算表情推荐环境信息特征向量H，H＝H1*Wh1+H2*Wh2+H3*Wh3+H4*Wh4+……Hn*Whn，其中Whn为各特征向量的权重。H的向量最终会结合以下其他特征计算出表情推荐向量T。
三、生理信息检测：通过摄像头获取当前用户脸部表情L1，可以通过连接可选外部可穿戴设备检测用户心跳L2、血压L3、温度L4等；根据以上特征向量可以计算表情推荐生理信息特征向量L，L＝L1*Wl1+L2*Wl2+L3*Wl3+L4*Wl4+……Ln*Wln，其中Wln为各特征向量的权重。L的向量最终会结合其他特征计算出表情推荐向量T。
四、推测模型：根据会话上下文检测、环境信息检测、生理信息检测提供的信息经过内部处理计算出表情推荐向量T，再更具向量T关联用户需要发送的可能的多个表情，并按优先级排序。T的计算表达式如下：
T＝S*W1+H*W2+L*W3
其中，W1为表情推荐会话上下文信息特征向量的权重；W2为表情推荐环境信息特征向量的权重；W3为表情推荐环境信息特征向量的权重，且W1+W2+W3＝1。
五、表情推荐：向用户按优先级呈现推荐的表情供选择；
六、历史数据记录：记录会话上下文检测、环境信息检测、生理信息检测的数据，供学习模型使用；
七、学习模型：每隔一个时间段依据历史数据记录训练模型学习用户在多个数据条件下表情发送的选择，训练完成后生成新的推测模型，替换原有推测模型。学习模型主要用于学习各权重值W，在用户首次使用表情推荐功能时，推荐模型将会采用一个通用的权重值集合进行初始化计算，如果推荐的表情用户并未采用，用户发送了非推荐的表情，学习模型将会学习用户这一行为，积累到一定时间段重新根据用户的行为进行模型的学习和训练新的模型更新W值以替换旧有推荐模型，在不断的学习和更新中逐步优化更加贴近用户的行为，提高推荐准确率
图4是本公开实施例提供的表情推荐的方法流程图，如图4所示，包括：S401：通过装置提供的能力获取及时通讯软件会话上下文信息、环境信息、生理信息；S402:推测模型根据以上信息，推测用户需要发送的表情；S403：表情推荐界面呈现推荐的表情供用户选择。
图5是本公开实施例提供的模型学习的方法流程图，如图5所示，包括：S501：记录会话上下文信息、环境信息、生理信息；S502：学习模型依据历史数据记录训练生成新的推测模型；S503：新的推测模型替换原有推测模型。
根据本公开实施例提供的方案，根据会话期间的会话上下文信息、环境信息、生理信息，从推测模型中可以自动向用户推荐表情，供用户选择，免去用户繁琐的表情查找步骤，并且通过自学习模型，可以不断改进和优化推测模型，以便更好地向用户推荐表情，提升了用户体验。
尽管上文对本公开进行了详细说明，但是本公开不限于此，本技术领域技术人员可以根据本公开的原理进行各种修改。因此，凡按照本公开原理所作的修改，都应当理解为落入本公开的保护范围。

Claims (10)

  1. 一种表情输入的方法，其中，应用于终端上，包括：
    在用户会话期间，获取所述用户会话期间的会话情景信息；
    根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；
    显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中。
  2. 根据权利要求1所述的方法，其中，所述会话情景信息至少包括以下任一：用户会话上下文信息、当前环境信息以及用户生理信息；
    其中，所述用户会话上下文信息至少包括以下任一：会话文本信息、会话场景信息、会话对象信息以及发送表情信息；所述当前环境信息至少包括以下任一：地理位置、日期、天气、温度以及终端的应用使用情况；所述用户生理信息至少包括以下任一：用户脸部表情、用户心跳、用户血压以及用户温度。
  3. 根据权利要求1所述的方法，其中，根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情包括：
    分别计算出所述会话情景信息中的用户会话上下文信息特征向量S、当前环境信息特征向量H以及用户生理信息特征向量L；
    根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T；
    根据所述表情推荐向量T，从表情推测模型中确定与所述表情推荐向量T相匹配的推荐表情。
  4. 根据权利要求3所述的方法，其中，所述根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T包括：
    T＝S*W1+H*W2+L*W3；
    其中，W1是所述用户会话上下文信息特征向量S的权重；所述W2是所述当前环境信息特征向量H的权重；所述W3是所述用户生理信息特征向量L的权重；且W1+W2+W3＝1。
  5. 根据权利要求1-4任一所述的方法，其中，还包括对所述表情推测模型进行更新的操作步骤，其包括：
    将所述会话情景信息及在确定推荐表情后用户所选取的推荐表情或非推荐表情保存到历史数据库中；
    当检测到需要对表情推测模型进行更新时，利用所述历史数据库对所述表情推测模型进行训练，生成新的表情推测模型；
    将所述新的表情推测模型进行保存，同时删除所述表情推测模型。
  6. 一种表情输入的装置，其中，包括：
    获取模块，用于在用户会话期间，获取所述用户会话期间的会话情景信息；
    推测模块，用于根据所述会话情景信息，从表情推测模型中确定与所述会话情景信息相匹配的推荐表情；
    显示模块，用于显示所述推荐表情，以便用户选取所述推荐表情后发送到所述用户会话中。
  7. 根据权利要求6所述的装置，其中，所述会话情景信息至少包括以下任一：用户会话上下文信息、当前环境信息以及用户生理信息；
    其中，所述用户会话上下文信息至少包括以下任一：会话文本信息、会话场景信息、会话对象信息以及发送表情信息；所述当前环境信息至少包括以下任一：地理位置、日期、天气、温度以及终端的应用使用情况；所述用户生理信息至少包括以下任一：用户脸部表情、用户心跳、用户血压以及用户温度。
  8. 根据权利要求6所述的装置，其中，所述推测模块包括：
    第一计算单元，用于分别计算出所述会话情景信息中的用户会话上下文信息特征向量S、当前环境信息特征向量H以及用户生理信息特征向量L；
    第二计算单元，用于根据所述用户会话上下文信息特征向量S、所述当前环境信息特征向量H以及所述用户生理信息特征向量L，计算出表情推荐向量T；
    推测单元，用于根据所述表情推荐向量T，从表情推测模型中确定与所述表情推荐向量T相匹配的推荐表情。
  9. 一种表情输入的设备，其中，所述设备包括：处理器，以及与所述处理器耦接的存储器；所述存储器上存储有可在所述处理器上运行的表情输入的程序， 所述表情输入的程序被所述处理器执行时实现如权利要求1至5中任一项所述的表情输入的方法的步骤。
  10. 一种计算机存储介质，其中，所述存储介质存储有表情输入的程序，所述表情输入的程序被处理器执行时实现如权利要求1至5中任一项所述的表情输入的方法的步骤。
PCT/CN2019/117846 2018-11-15 2019-11-13 一种表情输入的方法、装置、设备以及存储介质 WO2020098669A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811356978.6 2018-11-15
CN201811356978.6A CN111190493A (zh) 2018-11-15 2018-11-15 一种表情输入的方法、装置、设备以及存储介质

Publications (1)

Publication Number Publication Date
WO2020098669A1 true WO2020098669A1 (zh) 2020-05-22

Family

ID=70707043

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/117846 WO2020098669A1 (zh) 2018-11-15 2019-11-13 一种表情输入的方法、装置、设备以及存储介质

Country Status (2)

Country Link
CN (1) CN111190493A (zh)
WO (1) WO2020098669A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111726843A (zh) * 2020-05-29 2020-09-29 新华三技术有限公司成都分公司 一种建立会话的方法、一种设备、一种存储介质

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114511713A (zh) * 2022-04-20 2022-05-17 威海经济技术开发区天智创新技术研究院 一种基于图像的预测方法、装置及服务器

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050234727A1 (en) * 2001-07-03 2005-10-20 Leo Chiu Method and apparatus for adapting a voice extensible markup language-enabled voice system for natural speech recognition and system response
CN106293120A (zh) * 2016-07-29 2017-01-04 维沃移动通信有限公司 表情输入方法及移动终端
CN107634901A (zh) * 2017-09-19 2018-01-26 广东小天才科技有限公司 会话表情的推送方法、推送装置及终端设备
CN107729320A (zh) * 2017-10-19 2018-02-23 西北大学 一种基于时序分析用户会话情感趋势的表情符号推荐方法
CN107784114A (zh) * 2017-11-09 2018-03-09 广东欧珀移动通信有限公司 表情图像的推荐方法、装置、终端及存储介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106789543A (zh) * 2015-11-20 2017-05-31 腾讯科技（深圳）有限公司 会话中实现表情图像发送的方法和装置
CN107423277A (zh) * 2016-02-16 2017-12-01 中兴通讯股份有限公司 一种表情输入方法、装置及终端
CN107145270A (zh) * 2017-04-25 2017-09-08 北京小米移动软件有限公司 表情图标排序方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050234727A1 (en) * 2001-07-03 2005-10-20 Leo Chiu Method and apparatus for adapting a voice extensible markup language-enabled voice system for natural speech recognition and system response
CN106293120A (zh) * 2016-07-29 2017-01-04 维沃移动通信有限公司 表情输入方法及移动终端
CN107634901A (zh) * 2017-09-19 2018-01-26 广东小天才科技有限公司 会话表情的推送方法、推送装置及终端设备
CN107729320A (zh) * 2017-10-19 2018-02-23 西北大学 一种基于时序分析用户会话情感趋势的表情符号推荐方法
CN107784114A (zh) * 2017-11-09 2018-03-09 广东欧珀移动通信有限公司 表情图像的推荐方法、装置、终端及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111726843A (zh) * 2020-05-29 2020-09-29 新华三技术有限公司成都分公司 一种建立会话的方法、一种设备、一种存储介质
CN111726843B (zh) * 2020-05-29 2023-11-03 新华三技术有限公司成都分公司 一种建立会话的方法、一种设备、一种存储介质

Also Published As

Publication number Publication date
CN111190493A (zh) 2020-05-22

Similar Documents

Publication Publication Date Title
CN110869969B (zh) 用于在通信会话内生成个性化响应的虚拟助手
CN110892395B (zh) 提供增强的通信会话服务的虚拟助手
KR102050334B1 (ko) 언어 모델을 사용한, 메시지들에서 수신되는 이미지들에 대한 자동적인 제안 응답들
US10872112B2 (en) Automatic suggestions to share images
US8812417B2 (en) Hierarchical based sequencing machine learning model
WO2020005648A1 (en) Meeting preparation manager
US20180285827A1 (en) Distinguishing events of users for efficient service content distribution
JP5819572B1 (ja) 階層ベース配列の機械学習モデル
CN114556354A (zh) 自动确定和呈现来自事件的个性化动作项
CN110688528B (zh) 生成视频的分类信息的方法、装置、电子设备和介质
US20160104094A1 (en) Future meeting evaluation using implicit device feedback
KR20180070659A (ko) 메시징 애플리케이션들을 위한 자동적인 제안들 및 다른 콘텐츠
US20160019280A1 (en) Identifying question answerers in a question asking system
US10726087B2 (en) Machine learning system and method to identify and connect like-minded users
WO2017143773A1 (zh) 一种众包学习方法及装置
JP7488871B2 (ja) 対話推薦方法、装置、電子機器、記憶媒体ならびにコンピュータプログラム
WO2020098669A1 (zh) 一种表情输入的方法、装置、设备以及存储介质
EP4371027A1 (en) Intelligent task completion detection at a computing device
US20230388422A1 (en) Determination and display of estimated hold durations for calls
CN109615458A (zh) 客户管理方法、装置、终端设备及计算机可读存储介质
US20230325944A1 (en) Adaptive wellness collaborative media system
US11765274B2 (en) Determination and display of estimated hold durations for calls
CN116775815B (zh) 对话数据的处理方法、装置、电子设备及存储介质
WO2022156468A1 (zh) 模型数据处理方法、装置、电子设备以及计算机可读介质
CN115665363A (zh) 一种视频会议方法、装置、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: 19883817

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: 19883817

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 15/09/2021)