WO2020206957A1 - 一种应用于智能客服机器人的意图识别方法及装置 - Google Patents
一种应用于智能客服机器人的意图识别方法及装置 Download PDFInfo
- Publication number
- WO2020206957A1 WO2020206957A1 PCT/CN2019/109122 CN2019109122W WO2020206957A1 WO 2020206957 A1 WO2020206957 A1 WO 2020206957A1 CN 2019109122 W CN2019109122 W CN 2019109122W WO 2020206957 A1 WO2020206957 A1 WO 2020206957A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- text
- intention
- user
- word segmentation
- word
- Prior art date
Links
Images
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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
Definitions
- the invention relates to the field of artificial intelligence technology, and in particular to an intention recognition method and device applied to an intelligent customer service robot.
- customer service robots can effectively share the workload of manual customer service, save enterprise employment costs, break through time, manpower, and geographical constraints, and provide 7*24 continuous consulting services to ease Pain points of manual customer service.
- Customer service robots can accept various questions raised by users.
- One of the keys to the efficient use of customer service robots is whether they can determine the true intention of the user based on the information given by the user.
- the embodiments of the present invention provide an intention recognition method and device applied to an intelligent customer service robot, so as to realize the intelligent customer service robot quickly and accurately recognize the user's intention, and provide guarantee for the robot to accurately answer the user's question.
- an intention recognition method applied to an intelligent customer service robot includes the steps:
- step S2 Determine whether the dialogue text contains an intention, if it does, execute step S4, if it does not, end the process, if it cannot be judged, execute step S3;
- step S3 Perform context extension on the dialogue text, and after step S3, perform step S4;
- S5 The dialogue text is represented by a distributed word vector, and multiple pre-trained semantic classification models are used for prediction to obtain multiple semantic information;
- S6 Use the Ensemble framework to merge and optimize the intention knowledge points and the multiple semantic information to obtain user intentions.
- the method further includes the steps:
- step S1 specifically includes:
- step S3 specifically includes:
- the dialogue text is expanded by using synonyms of the context.
- step S4 specifically includes:
- step S5 specifically includes:
- step S6 specifically includes:
- the final user intention is determined through the Ensemble framework.
- an intent recognition device applied to an intelligent customer service robot includes:
- the text acquisition module is used to acquire the dialog text of the user
- the intention judgment module is used to judge whether the dialogue text contains an intention, if it does, execute the processing of the entity matching module, if it does not, end the processing, if it cannot be judged, execute the processing of the text expansion module;
- the text expansion module is configured to expand the context of the dialog text, and execute the processing of the entity matching module for the expanded dialog text;
- the entity matching module is used to identify the set of named entities in the dialogue text and determine the intent knowledge points associated with the set of named entities;
- the semantic prediction module is used to represent the dialogue text using distributed word vectors, and use a plurality of pre-trained semantic classification models for prediction to obtain a plurality of semantic information;
- the merging and tuning module is used to merge and tune the intention knowledge points and the multiple semantic information using the Ensemble framework to obtain the user's intention.
- the device further includes:
- the device also includes:
- the text error correction module is used to perform text error correction on the dialog text.
- text error correction module is specifically used for:
- text extension module is specifically used for:
- the dialogue text is expanded by using synonyms of the context.
- entity matching module is specifically configured to:
- semantic prediction module is specifically used for:
- merge tuning module is specifically used for:
- the final user intention is determined through the Ensemble framework.
- the present invention has the following beneficial effects:
- Figure 1 shows a flow chart of an intention recognition method applied to an intelligent customer service robot
- FIG. 2 shows a specific implementation flow chart of step S1 in Figure 1;
- FIG. 3 shows a specific implementation flow chart of step S3 in Figure 1;
- FIG. 4 shows a specific implementation flowchart of step S4 in FIG. 1;
- FIG. 5 shows a specific implementation flow chart of step S5 in Figure 1;
- Fig. 6 shows a block diagram of an intention recognition device applied to an intelligent customer service robot.
- the embodiment of the present invention provides an intention recognition method applied to an intelligent customer service robot.
- the method expands the context of the conversation text and combines entity matching recognition and semantic information prediction to obtain user intentions, which can identify users more quickly and accurately Intent, improves the accuracy of recognizing user intent, reduces the error and incompleteness of user intent recognition, thereby providing guarantee for customer service robots to correctly answer user questions.
- any intelligent terminal which includes but is not limited to a desktop computer, a personal computer, a smart phone, a tablet computer, and so on.
- the embodiment of the present invention provides an intention recognition method applied to an intelligent customer service robot. As shown in FIG. 1, the method includes the following steps:
- the user dialogue may be voice or text.
- the user dialogue may be converted from voice to text before the implementation of the embodiment of the present invention.
- the dialog text may be a long text or a short text, which is not specifically limited in the embodiment of the present invention.
- step S1 may include:
- S11 Perform word segmentation on the dialogue text and identify the wrong word segmentation in the dialogue text.
- the error correction words corresponding to the error segmentation can be obtained based on the typos dictionary. Specifically, for the error segmentation, the error correction confidence level corresponding to each word in the custom standard dictionary is calculated, and the error correction confidence level is greater than the preset Threshold words are used as error correction words.
- edit distance or language model may also be used to obtain the error correction words corresponding to the error segmentation, and the specific obtaining process in this embodiment is not specifically limited.
- the error correction words are mainly used to correct the wrong word segmentation in the recognized text. For example, if a wrong word segmentation in the recognized text "Big Tree Data Application Case” is “Big Tree Data”, the corresponding error correction word is "Big Data”.
- step S1 is an optional process.
- the dialog text with incorrect words is converted into a correct expression that conforms to the domain logic, so that the user's intention can be more accurately recognized.
- step S2 Determine whether the dialogue text contains an intention, if it is included, then perform step S4, if it does not, then end the process, if it cannot be determined, then perform step S3.
- step S2 may include:
- the preset template can adopt regular expressions. Style pattern.
- the text expressed by the user in the customer service robot may be a dialogue text with only a few words, the user's expression is very ambiguous.
- the process of step S2 may not be able to determine whether the user's dialogue contains intent, the dialogue text needs to be contextualized Extension.
- step S3 Perform context extension on the dialogue text, and after step S3, perform step S4.
- step S3 may include:
- S31 Save the user's session information in a unit of session, contact the context information of the dialog text, and determine whether the user's intention has changed, where the context information includes the contextual intention recognition result of the dialog text.
- the relevant information related to the context can be used, with a session as a unit, using the conversation information saved in a session, and combining multiple conversation texts input by the user before to determine whether the intention has changed.
- the keywords in the context are extracted to obtain a set of synonyms, and the set of synonyms is used to expand the dialogue text.
- the intention information in the dialog text can be enriched, so that the user's intention can be accurately identified in the subsequent.
- step S4 may include:
- S41 Perform word segmentation processing on the dialogue text according to a preset dictionary to obtain multiple word segmentation.
- the dialogue text is segmented using a preset word segmentation method to obtain multiple characters or character sequences, and characters or character sequences with actual semantics are selected from the obtained character sequences as the word segmentation result.
- the preset word segmentation method may be a word segmentation method based on character matching, semantic understanding, or statistics.
- the matching degree between each named entity in the entity dictionary and the word segmentation is calculated, and the named entity with the matching degree greater than the preset threshold is used as the name matching the word segmentation entity.
- the similarity based on Hamming distance can be used to calculate the matching degree between each named entity in the entity vocabulary and the word segmentation.
- S43 Determine intent knowledge points related to the named entity set in the preset knowledge base.
- multiple entities correspond to one intention knowledge point
- the intention knowledge point is used to refer to schematic diagram information.
- the standardized intention knowledge points can be collected and sorted out in advance.
- Each intent knowledge point determines the corresponding multiple entities.
- a preliminary prediction of user intent can be obtained.
- the correlation between the named entity set and each intention knowledge point in the knowledge base is calculated, and the intention knowledge point associated with the named entity set in the knowledge base is determined.
- S5 The dialogue text is represented by distributed word vectors, and multiple pre-trained semantic classification models are used for prediction to obtain multiple semantic information.
- step S5 may include:
- S51 Perform word segmentation processing on the dialogue text to obtain multiple word segments.
- step S41 the specific process of this step is the same as that of step S41, and will not be repeated here.
- S52 Calculate the word vectors of multiple word segments, and distribute the word vectors of the multiple word segments.
- the word vector corresponding to the word unit can be obtained through the Word2Vec model, and the distributed representation of the word vector can be performed.
- Word2Vec is a specific method of word embedding for natural language processing NLP. It can represent the semantic information of words by learning text by word vector, that is, make semantically similar words in an embedding space (low-dimensional). The distance in this space is very close.
- S53 Input the word vectors of the multiple segmented words in a distributed representation to multiple semantic classification models to output multiple semantic information.
- step S5 the training process of multiple semantic classification models in step S5 respectively includes the following:
- the question and answer data includes the question and answer pair information accumulated by the field human customer service when answering user questions.
- keyword extraction and template rules can be used to preprocess the question and answer data to filter out part of the unintentional data, and the preprocessed question and answer data can be semantically annotated by annotators.
- the semantic classification in the field can be subdivided into multiple classifications including phone bills, gift cards, wealth management, change treasure, etc., and the question and answer data are pre-labeled by the labelers.
- the above-mentioned multiple semantic classification models can use a variety of deep learning semantic classification models such as TextCNN, RNN, LSTM, CAPSNet, etc.
- the model training strategy can use the conventional strategy of the corresponding network. No longer describe.
- test set After training the deep neural network using the training set, you can use the test set to test multiple deep neural networks after training to evaluate the prediction accuracy of the deep neural network, and adjust the network parameters of the deep neural network according to the model prediction accuracy , In order to construct a semantic classification model whose prediction accuracy meets the accuracy threshold.
- the semantic classification of the question and answer data is represented by word vector distribution
- the deep learning model is used for deep feature mining, and the semantic relationship between words is fully considered while extracting the features to obtain the semantic classification model. Therefore, multiple semantic classification models can be used to quickly and accurately predict the semantic information of the user dialogue text represented by the word vector distribution.
- S6 Use the Ensemble framework to merge and optimize intention knowledge points and multiple semantic information to obtain user intentions.
- the final user intention is determined through the Ensemble framework.
- the basic idea of the Ensemble framework is to make full use of the various advantages of different classification algorithms, learn from each other, and combine to form a powerful classification framework. Combine the results of multiple classifiers to achieve the best combination.
- the intent recognition method applied to the intelligent customer service robot provided by the embodiment of the present invention can complete the intent information in the user’s dialog text by expanding the context of the dialog text when it is impossible to determine whether the dialog text contains the intention; Distributed word vector representation and deep learning model for deep feature mining, while extracting features, fully consider the semantic association between words; through the use of the Ensemble framework, the entity matching results and semantic prediction results are combined and optimized to obtain user intentions and achieve more In order to quickly and accurately identify the user's intention, the accuracy of identifying the user's intention is improved, and the error and incompleteness of the user's intention identification are reduced, thereby providing a guarantee for the customer service robot to answer the user's question correctly.
- an embodiment of the present invention provides an intention recognition device applied to an intelligent customer service robot. As shown in FIG. 6, the device includes:
- the text acquisition module 60 is used to acquire the dialog text of the user
- the intention judgment module 62 is used to judge whether an intention is included in the dialog text, if it is included, then execute the processing of the entity matching module, if it is not included, end the processing, if it cannot be judged, execute the processing of the text expansion module 63;
- the text expansion module 63 is used to expand the context of the dialog text, and execute the processing of the entity matching module for the expanded dialog text;
- the entity matching module 64 is used to identify the named entity set in the dialogue text and determine the intention knowledge points associated with the named entity set;
- the semantic prediction module 65 is used to represent the dialogue text using distributed word vectors, and use multiple pre-trained semantic classification models for prediction to obtain multiple semantic information;
- the merging and tuning module 66 is used to merge and tune intent knowledge points and semantic information using the Ensemble framework to obtain user intent.
- the device also includes:
- the text error correction module 61 is used to perform text error correction on the dialogue text.
- the text error correction module 61 is specifically used for:
- the text extension module 63 is specifically used for:
- entity matching module 64 is specifically configured to:
- semantic prediction module 65 is specifically used for:
- merging and tuning module 66 is specifically used for:
- the final user intention is determined through the Ensemble framework.
- the intent recognition learning device applied to the intelligent customer service robot provided in this embodiment belongs to the same inventive concept as the intent recognition method applied to the intelligent customer service robot provided in the embodiment of the present invention, and can perform the application provided by any embodiment of the present invention.
- the intent recognition method of the intelligent customer service robot has functional modules and beneficial effects that execute the corresponding intent recognition method applied to the intelligent customer service robot. For technical details that are not described in detail in this embodiment, please refer to the intent recognition method applied to the intelligent customer service robot provided in the embodiment of the present invention, which will not be repeated here.
- a person of ordinary skill in the art can understand that all or part of the steps in the above embodiments can be implemented by hardware, or by a program instructing associated hardware to complete, and the program can be stored in a computer-readable storage medium.
- the aforementioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
Abstract
Description
Claims (14)
- 一种应用于智能客服机器人的意图识别方法,其特征在于,包括步骤:S0:获取用户的对话文本;S2:判断所述对话文本中是否包含意图,若包含,则执行步骤S4,若未包含,则结束处理,若无法判断,则执行步骤S3;S3:将所述对话文本进行上下文扩展,在步骤S3之后,执行步骤S4;S4:识别所述对话文本中的命名实体集合,并确定所述命名实体集合关联的意图知识点;S5:将所述对话文本采用分布式词向量进行表示,并使用预先训练的多个语义分类模型进行预测,得到多个语义信息;S6:使用Ensemble框架对所述意图知识点和所述多个语义信息进行合并调优,得到用户意图。
- 根据权利要求1所述的方法,其特征在于,所述步骤S2之前,所述方法还包括步骤:S1:对所述对话文本进行文本纠错。
- 根据权利要求2所述的方法,其特征在于,所述步骤S1具体包括:对所述对话文本进行分词,并识别所述对话文本中的错误分词;获取所述错误分词对应的纠错词;将所述纠错词替换所述对话文本中的错误分词。
- 根据权利要求1至3任意一项所述的方法,其特征在于,所述步骤S3具体包括:以一个session为单位保存用户会话信息;联系所述对话文本的上下文信息,判断用户意图是否改变,其中,所述上下文信息包括所述对话文本的上下文的意图识别结果;当用户意图未改变时,利用所述上下文的近义词对所述对话文本进行扩展。
- 根据权利要求1至3任意一项所述的方法,其特征在于,所述步骤S4具体包括:根据预设的词典对所述对话文本进行分词处理,得到多个分词;将所述多个分词与预设的实体词库进行匹配,得到所述命名实体集合;在预设的知识库中确定出与所述命名实体集合相关的意图知识点。
- 根据权利要求1至3任意一项所述的方法,其特征在于,所述步骤S5具体包括:将所述对话文本进行分词处理,得到多个分词;计算所述多个分词的词向量,并分布式表示所述多个分词的词向量;将分布式表示的所述多个分词的词向量输入至所述多个语义分类模型,以输出所述多个语义信息。
- 根据权利要求1至3任意一项所述的方法,其特征在于,所述步骤S6具体包括:根据所述意图知识点、所述多个语义信息以及所述多个语义信息各自对应的预设权重,通过Ensemble框架确定最终的用户意图。
- 一种应用于智能客服机器人的意图识别装置,其特征在于,包括:文本获取模块,用于获取用户的对话文本;意图判断模块,用于判断所述对话文本中是否包含意图,若包括,则执行实体匹配模块的处理,若未包含,则结束处理,若无法判断,则执行文本扩展模块的处理;所述文本扩展模块,用于将所述对话文本进行上下文扩展,并针对扩展后的所述对话文本执行实体匹配模块的处理;所述实体匹配模块,用于识别所述对话文本中的命名实体集合,并确定所述命名实体集合关联的意图知识点;语义预测模块,用于将所述对话文本采用分布式词向量进行表示,并使用预先训练的多个语义分类模型进行预测,得到多个语义信息;合并调优模块,用于使用Ensemble框架对所述意图知识点和所述多个语义信息进行合并调优,得到用户意图。
- 根据权利要求8所述的装置,其特征在于,所述装置还包括:文本纠错模块,用于对所述对话文本进行文本纠错。
- 根据权利要求9所述的装置,其特征在于,所述文本纠错模块具体用于:对所述对话文本进行分词,并识别所述对话文本中的错误分词;获取所述错误分词对应的纠错词;将所述纠错词替换所述对话文本中的错误分词。
- 根据权利要求8至10任意一项所述的装置,其特征在于,所述文本扩展模块具体用于:以一个session为单位保存用户会话信息;联系所述对话文本的上下文信息,判断用户意图是否改变,其中,所述上下文信息包括所述对话文本的上下文的意图识别结果;当用户意图未改变时,利用所述上下文的近义词对所述对话文本进行扩展。
- 根据权利要求8至10任意一项所述的装置,其特征在于,所述实体匹配模块具体用于:根据预设的词典对所述对话文本进行分词处理,得到多个分词;将所述多个分词与预设的实体词库进行匹配,得到所述命名实体集合;在预设的知识库中确定出与所述命名实体集合相关的意图知识点。
- 根据权利要求8至10任意一项所述的装置,其特征在于,所述语义预测模块具体用于:将所述对话文本进行分词处理,得到多个分词;计算所述多个分词的词向量,并分布式表示所述多个分词的词向量;将分布式表示的所述多个分词的词向量输入至所述多个语义分类模型,以输出所述多个语义信息。
- 根据权利要求8至10任意一项所述的装置,其特征在于,所述合并调优模块具体用于:根据所述意图知识点、所述多个语义信息以及所述多个语义信息各自对应的预设权重,通过Ensemble框架确定最终的用户意图。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CA3176868A CA3176868A1 (en) | 2019-04-09 | 2019-09-29 | Intent identifying method and device for application to intelligent customer service robot |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910281032.6A CN110096570B (zh) | 2019-04-09 | 2019-04-09 | 一种应用于智能客服机器人的意图识别方法及装置 |
CN201910281032.6 | 2019-04-09 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020206957A1 true WO2020206957A1 (zh) | 2020-10-15 |
Family
ID=67444578
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/109122 WO2020206957A1 (zh) | 2019-04-09 | 2019-09-29 | 一种应用于智能客服机器人的意图识别方法及装置 |
Country Status (3)
Country | Link |
---|---|
CN (1) | CN110096570B (zh) |
CA (1) | CA3176868A1 (zh) |
WO (1) | WO2020206957A1 (zh) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112364622A (zh) * | 2020-11-11 | 2021-02-12 | 杭州大搜车汽车服务有限公司 | 对话文本分析方法、装置、电子装置及存储介质 |
CN112487827A (zh) * | 2020-12-28 | 2021-03-12 | 科大讯飞华南人工智能研究院(广州)有限公司 | 问题回答方法及电子设备、存储装置 |
CN112541792A (zh) * | 2020-12-22 | 2021-03-23 | 作业帮教育科技(北京)有限公司 | 一种挖掘用户需求的数据处理方法、装置及电子设备 |
Families Citing this family (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110096570B (zh) * | 2019-04-09 | 2021-03-30 | 苏宁易购集团股份有限公司 | 一种应用于智能客服机器人的意图识别方法及装置 |
CN110457709A (zh) * | 2019-08-16 | 2019-11-15 | 北京一链数云科技有限公司 | 外呼对话处理方法、装置及服务器 |
CN110569331A (zh) * | 2019-09-04 | 2019-12-13 | 出门问问信息科技有限公司 | 一种基于上下文的关联性预测方法、装置及存储设备 |
CN112487179A (zh) * | 2019-09-11 | 2021-03-12 | 珠海格力电器股份有限公司 | 一种口语语义理解方法、装置及系统 |
CN110717026B (zh) * | 2019-10-08 | 2023-09-22 | 腾讯科技(深圳)有限公司 | 一种文本信息识别的方法、人机对话的方法及相关装置 |
CN110888968A (zh) * | 2019-10-15 | 2020-03-17 | 浙江省北大信息技术高等研究院 | 客服对话意图分类方法及装置、电子设备及介质 |
CN112668333A (zh) * | 2019-10-15 | 2021-04-16 | 华为技术有限公司 | 命名实体的识别方法和设备、以及计算机可读存储介质 |
CN111026843B (zh) * | 2019-12-02 | 2023-03-14 | 北京智乐瑟维科技有限公司 | 一种人工智能语音外呼方法、系统及存储介质 |
CN111091826B (zh) * | 2019-12-13 | 2023-09-01 | 中博信息技术研究院有限公司 | 基于深度学习和有限状态机的智能语音机器人系统 |
CN111078855A (zh) * | 2019-12-19 | 2020-04-28 | 联想(北京)有限公司 | 信息处理方法、装置、电子设备及存储介质 |
CN111160002B (zh) * | 2019-12-27 | 2022-03-01 | 北京百度网讯科技有限公司 | 用于输出口语理解中解析异常信息的方法和装置 |
CN111325037B (zh) * | 2020-03-05 | 2022-03-29 | 苏宁云计算有限公司 | 文本意图识别方法、装置、计算机设备和存储介质 |
CN111462752B (zh) * | 2020-04-01 | 2023-10-13 | 北京思特奇信息技术股份有限公司 | 基于注意力机制、特征嵌入及bi-lstm的客户意图识别方法 |
CN111460122A (zh) * | 2020-04-03 | 2020-07-28 | 成都晓多科技有限公司 | 基于深度学习的尺码识别方法与系统 |
CN112148862B (zh) * | 2020-10-15 | 2024-01-30 | 腾讯科技(深圳)有限公司 | 一种问题意图识别方法、装置、存储介质及电子设备 |
CN112256854A (zh) * | 2020-11-05 | 2021-01-22 | 云南电网有限责任公司 | 一种基于ai自然语言理解的智能ai会话方法及装置 |
CN112562665A (zh) * | 2020-11-30 | 2021-03-26 | 武汉海昌信息技术有限公司 | 一种基于信息交互的语音识别方法、存储介质及系统 |
CN112364149B (zh) * | 2021-01-12 | 2021-04-23 | 广州云趣信息科技有限公司 | 用户问题获得方法、装置及电子设备 |
CN113076403A (zh) * | 2021-04-21 | 2021-07-06 | 深圳追一科技有限公司 | 一种用户消息处理方法及相关设备 |
CN113282737B (zh) * | 2021-07-21 | 2021-11-12 | 中信建投证券股份有限公司 | 人机协作的智能客服对话方法及装置 |
CN113569578B (zh) * | 2021-08-13 | 2024-03-08 | 上海淇玥信息技术有限公司 | 一种用户意图识别方法、装置和计算机设备 |
CN114118080B (zh) * | 2021-11-10 | 2022-09-13 | 北京深维智信科技有限公司 | 一种从销售会话中自动识别客户意向的方法及系统 |
CN114118060B (zh) * | 2021-11-10 | 2022-09-27 | 北京深维智信科技有限公司 | 一种从销售会话中自动识别关键事件的方法及系统 |
CN114706945A (zh) * | 2022-03-24 | 2022-07-05 | 马上消费金融股份有限公司 | 意图识别方法、装置、电子设备及存储介质 |
CN115130465A (zh) * | 2022-07-18 | 2022-09-30 | 浙大城市学院 | 文献数据集上知识图谱实体标注错误识别方法和系统 |
CN116468024B (zh) * | 2023-04-13 | 2023-09-29 | 重庆明度科技有限责任公司 | Ai上下文生成方法 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180052885A1 (en) * | 2016-08-16 | 2018-02-22 | Ebay Inc. | Generating next user prompts in an intelligent online personal assistant multi-turn dialog |
CN108363690A (zh) * | 2018-02-08 | 2018-08-03 | 北京十三科技有限公司 | 基于神经网络的对话语义意图预测方法及学习训练方法 |
CN108829757A (zh) * | 2018-05-28 | 2018-11-16 | 广州麦优网络科技有限公司 | 一种聊天机器人的智能服务方法、服务器及存储介质 |
CN109461039A (zh) * | 2018-08-28 | 2019-03-12 | 厦门快商通信息技术有限公司 | 一种文本处理方法及智能客服方法 |
CN110096570A (zh) * | 2019-04-09 | 2019-08-06 | 苏宁易购集团股份有限公司 | 一种应用于智能客服机器人的意图识别方法及装置 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20110036385A (ko) * | 2009-10-01 | 2011-04-07 | 삼성전자주식회사 | 사용자 의도 분석 장치 및 방법 |
US9183310B2 (en) * | 2012-06-12 | 2015-11-10 | Microsoft Technology Licensing, Llc | Disambiguating intents within search engine result pages |
CN105487663B (zh) * | 2015-11-30 | 2018-09-11 | 北京光年无限科技有限公司 | 一种面向智能机器人的意图识别方法和系统 |
CN107193865B (zh) * | 2017-04-06 | 2020-03-10 | 上海奔影网络科技有限公司 | 人机交互中自然语言意图理解方法及装置 |
CN108763510B (zh) * | 2018-05-30 | 2021-10-15 | 北京五八信息技术有限公司 | 意图识别方法、装置、设备及存储介质 |
CN108874782B (zh) * | 2018-06-29 | 2019-04-26 | 北京寻领科技有限公司 | 一种层次注意力lstm和知识图谱的多轮对话管理方法 |
CN109241251B (zh) * | 2018-07-27 | 2022-05-27 | 众安信息技术服务有限公司 | 一种会话交互方法 |
-
2019
- 2019-04-09 CN CN201910281032.6A patent/CN110096570B/zh active Active
- 2019-09-29 WO PCT/CN2019/109122 patent/WO2020206957A1/zh active Application Filing
- 2019-09-29 CA CA3176868A patent/CA3176868A1/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180052885A1 (en) * | 2016-08-16 | 2018-02-22 | Ebay Inc. | Generating next user prompts in an intelligent online personal assistant multi-turn dialog |
CN108363690A (zh) * | 2018-02-08 | 2018-08-03 | 北京十三科技有限公司 | 基于神经网络的对话语义意图预测方法及学习训练方法 |
CN108829757A (zh) * | 2018-05-28 | 2018-11-16 | 广州麦优网络科技有限公司 | 一种聊天机器人的智能服务方法、服务器及存储介质 |
CN109461039A (zh) * | 2018-08-28 | 2019-03-12 | 厦门快商通信息技术有限公司 | 一种文本处理方法及智能客服方法 |
CN110096570A (zh) * | 2019-04-09 | 2019-08-06 | 苏宁易购集团股份有限公司 | 一种应用于智能客服机器人的意图识别方法及装置 |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112364622A (zh) * | 2020-11-11 | 2021-02-12 | 杭州大搜车汽车服务有限公司 | 对话文本分析方法、装置、电子装置及存储介质 |
CN112541792A (zh) * | 2020-12-22 | 2021-03-23 | 作业帮教育科技(北京)有限公司 | 一种挖掘用户需求的数据处理方法、装置及电子设备 |
CN112487827A (zh) * | 2020-12-28 | 2021-03-12 | 科大讯飞华南人工智能研究院(广州)有限公司 | 问题回答方法及电子设备、存储装置 |
Also Published As
Publication number | Publication date |
---|---|
CA3176868A1 (en) | 2020-10-15 |
CN110096570A (zh) | 2019-08-06 |
CN110096570B (zh) | 2021-03-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020206957A1 (zh) | 一种应用于智能客服机器人的意图识别方法及装置 | |
CN105869634B (zh) | 一种基于领域的带反馈语音识别后文本纠错方法及系统 | |
US11113323B2 (en) | Answer selection using a compare-aggregate model with language model and condensed similarity information from latent clustering | |
CN104050160B (zh) | 一种机器与人工翻译相融合的口语翻译方法和装置 | |
CN111241237B (zh) | 一种基于运维业务的智能问答数据处理方法及装置 | |
CN109857846B (zh) | 用户问句与知识点的匹配方法和装置 | |
CN110347787B (zh) | 一种基于ai辅助面试场景的面试方法、装置及终端设备 | |
CN111933127A (zh) | 一种具备自学习能力的意图识别方法及意图识别系统 | |
CN113094578A (zh) | 基于深度学习的内容推荐方法、装置、设备及存储介质 | |
CN111462752B (zh) | 基于注意力机制、特征嵌入及bi-lstm的客户意图识别方法 | |
CN112052319B (zh) | 一种基于多特征融合的智能客服方法及系统 | |
CN115617955A (zh) | 分级预测模型训练方法、标点符号恢复方法及装置 | |
US20230350929A1 (en) | Method and system for generating intent responses through virtual agents | |
CN113705207A (zh) | 语法错误识别方法及装置 | |
CN113157887A (zh) | 知识问答意图识别方法、装置、及计算机设备 | |
CN115905187B (zh) | 一种面向云计算工程技术人员认证的智能化命题系统 | |
CN110377753B (zh) | 基于关系触发词与gru模型的关系抽取方法及装置 | |
CN112349294A (zh) | 语音处理方法及装置、计算机可读介质、电子设备 | |
TW202034207A (zh) | 使用意圖偵測集成學習之對話系統及其方法 | |
CN116090450A (zh) | 一种文本处理方法及计算设备 | |
CN114117069A (zh) | 一种用于知识图谱智能问答的语义理解方法及系统 | |
CN114239555A (zh) | 一种关键词提取模型的训练方法及相关装置 | |
CN114780786B (zh) | 一种基于瓶颈特征和残差网络的语音关键词检索方法 | |
Pan | Design of Foreign Language Intelligent Translation Recognition System Based on Improved GLR Algorithm | |
US11934794B1 (en) | Systems and methods for algorithmically orchestrating conversational dialogue transitions within an automated conversational system |
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: 19923872 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: 19923872 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 1205 DATED 22/04/2022) |
|
ENP | Entry into the national phase |
Ref document number: 3176868 Country of ref document: CA |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19923872 Country of ref document: EP Kind code of ref document: A1 |