The invention discloses a voice interaction method, and the method comprises the steps: obtaining initial training data, determining a plurality of label reasoning results of a sample in the initial training data according to a first preset model, determining a reference sample in the initial training data according to the plurality of label reasoning results of the sample, and selecting a target sample which is similar to the reference sample in semantics from the candidate training data, determining target training data, training a second preset model by using the target training data to obtain a voice interaction model, and completing voice interaction according to the voice interaction model. Thus, the target sample in the training data is selected to determine the target training data, the voice interaction model is trained according to the target training data, and the training duration is guaranteed. According to the method, the reference sample is determined according to the sample reasoning result, the target sample with semantics similar to that of the reference sample is selected, the model reasoning condition is considered in selection of the target sample, the credibility and effectiveness are guaranteed, and the prediction precision of the voice interaction model is guaranteed.

    本申请公开一种语音交互方法,所述方法包括:获取初始训练数据,根据第一预设模型确定初始训练数据中样本的多个标签推理结果,根据样本的多个标签推理结果确定初始训练数据中参考样本,在候选训练数据中选取与参考样本语义相近的目标样本,确定目标训练数据,利用目标训练数据训练第二预设模型以得到语音交互模型,根据语音交互模型完成语音交互。如此,本申请选取选训练数据中目标样本以确定目标训练数据,根据目标训练数据训练语音交互模型,训练时长得到保障。本申请根据样本推理结果确定参考样本,选取与参考样本语义相近的目标样本,目标样本的选取考虑到模型推理情况,可信度和有效性得到保障,语音交互模型的预测精度得到保障。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Voice interaction method, server and computer readable storage medium


    Weitere Titelangaben:

    语音交互方法、服务器及计算机可读存储介质


    Beteiligte:
    WANG YI (Autor:in)

    Erscheinungsdatum :

    2024-02-02


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G10L Analyse oder Synthese von Sprache , SPEECH ANALYSIS OR SYNTHESIS / B60R Fahrzeuge, Fahrzeugausstattung oder Fahrzeugteile, soweit nicht anderweitig vorgesehen , VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR



    Voice interaction method, server and computer readable storage medium

    FAN JUNFENG / ZHAO QUN / JIAN TAOWEN et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    Voice interaction method, server and computer readable storage medium

    HU JING | Europäisches Patentamt | 2024

    Freier Zugriff

    Voice interaction method, server and computer readable storage medium

    FAN JUNFENG / DING PENGJIE / ZHAO QUN et al. | Europäisches Patentamt | 2022

    Freier Zugriff

    VOICE INTERACTION METHOD, SERVER, AND COMPUTER-READABLE STORAGE MEDIUM

    ZHANG XIKANG / ZHAO YAO / WANG TIANYI | Europäisches Patentamt | 2024

    Freier Zugriff

    Voice interaction method, server and computer readable storage medium

    FAN JUNFENG / NING HONGKE / ZHAO QUN | Europäisches Patentamt | 2022

    Freier Zugriff