WO2020139865A1 - Systèmes et procédés pour des conversations automatisées améliorées - Google Patents
Systèmes et procédés pour des conversations automatisées améliorées Download PDFInfo
- Publication number
- WO2020139865A1 WO2020139865A1 PCT/US2019/068421 US2019068421W WO2020139865A1 WO 2020139865 A1 WO2020139865 A1 WO 2020139865A1 US 2019068421 W US2019068421 W US 2019068421W WO 2020139865 A1 WO2020139865 A1 WO 2020139865A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- message
- meaning
- conversation
- intent
- intents
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/041—Abduction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Definitions
- AI Artificial Intelligence
- systems and methods are provided for parsing a message in a conversation series. This involves receiving a message, which is a collection of a series of exchanges including earlier exchanges and a current exchange. The different exchanges are identified in the message, and the earlier ones are removed to isolate the current exchange. The current exchange is then divided up into sentences, and the language being used is detected. Language detection may include running different language models on the message, and selecting the language for the model with the highest confidence.
- the classification text is provided to sets of models for parallel prediction of the intent(s) of the message.
- Each model is predicts the intent with a given confidence level.
- Models are queried for based upon series of the conversation, the industry involved, the client the model is for, and the message campaign to select which of the machine learning models to use for intent determination. If a speech act was identified previously, the models used to analyze the statement may differ based upon which speech type exists. For example, some models may be well tuned for analyzing statements, but fail to have accurate results for questions.
- Mapping rules and/or prediction machine learning models are used to convert the intents into meanings. These meanings may be filtered (e.g., by a threshold cutoff of some other filtering methodology) to arrive at the final meaning of the exchange. It is also possible to apply a decision engine policy for the determination of the meaning. The decision engine policy removes at least one of the plurality of machine learning models and replaces a probability of the predicted meaning for the removed models to 1.
- the training of the AI system may be enabled by, or supplemented with, conventional CRM data.
- CRM data may be particularly useful when used to augment traditional training sets, and input from the training desk.
- social media exchanges may likewise be useful as a training source for the AI models. For example, a business often engages directly with customers on social media, leading to conversations back and forth that are again, specific and accurate to the business. As such this data may also be beneficial as a source of training material.
- Feedback may be collected from the conversational exchanges, in many
- the message receiver 530 may include a smart parser 531 which consumes the raw message and splits it into multiple portions, including differentiating between the salutation, reply, close, signature and other message components. This is performed by the search and removal of the outgoing message (in an email this is typically appended below a response message and should mirror the message sent originally to the contact), followed by pre-processing of the test to build features. Subsequently predictions are requested by models for what message segment a particular line of text belongs to, and the response is logically built.
- a smart parser 531 which consumes the raw message and splits it into multiple portions, including differentiating between the salutation, reply, close, signature and other message components. This is performed by the search and removal of the outgoing message (in an email this is typically appended below a response message and should mirror the message sent originally to the contact), followed by pre-processing of the test to build features. Subsequently predictions are requested by models for what message segment a particular line of text belongs to, and the response is logical
- the input for the smart parser may include raw email text (or other message information converted into a raw text format).
- the input into the smart parser function may include:
- the entire data processor 540 operates as a step function responsible for connecting all of the lambda functions.
- the step function may start with the smart message parser, proceed to message normalization, then language detection, and then data processing, which includes its own stepwise sets of functions (e.g., critical intent checking, model query, parallel prediction, sorting, meaning mapping, etc.).
- the message templates in the conversation are generated (at 820). If the series is populated (at 830), then the conversation is reviewed and submitted (at 840). Otherwise, the next message in the template is generated (at 820).
- Figure 9 provides greater details of an example of this sub-process for generating message templates. Initially the user is queried if an existing conversation can be leveraged for templates, or whether a new template is desired (at 910).
- the system may detect the language (at 1320) of the response using multiple language models and selecting the model with the highest confidence level.
- the sentences are tokenized (at 1330) which divides the response into separate sentences. This is performed because generally each sentence of a conversation includes a separate/discrete idea or intention. By separating each sentence the risk of token
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Machine Translation (AREA)
Abstract
L'invention concerne des systèmes et des procédés d'analyse d'un message dans une série de conversations. Ceci consiste à recevoir un message, à isoler l'échange actuel, à le diviser en phrases et à détecter la langue utilisée. Les phrases du message sont normalisées et tout « acte de parole » est identifié. De même, toute « intention critique » est identifiée. S'il n'y a pas d'intention critique, le texte de classification est fourni à des ensembles de modèles pour une prédiction parallèle des une ou plusieurs intentions du message. Des modèles sont interrogés en fonction d'une série de la conversation, de l'industrie impliquée, du client auquel est destiné le modèle, de la campagne de messagerie et de tout acte de parole présent. Des règles de mappage et/ou des modèles d'apprentissage machine de prédiction sont utilisés pour convertir les intentions en significations, qui sont filtrées. Il est également possible d'appliquer une politique de moteur de décision pour la détermination de la signification. Ceci est suivi par une extraction d'entité et une génération de réponse par mappage des significations à des actions.
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US201862784696P | 2018-12-24 | 2018-12-24 | |
US62/784,696 | 2018-12-24 | ||
US16/723,735 | 2019-12-20 | ||
US16/723,735 US20200143115A1 (en) | 2015-01-23 | 2019-12-20 | Systems and methods for improved automated conversations |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020139865A1 true WO2020139865A1 (fr) | 2020-07-02 |
Family
ID=71129898
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US2019/068421 WO2020139865A1 (fr) | 2018-12-24 | 2019-12-23 | Systèmes et procédés pour des conversations automatisées améliorées |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2020139865A1 (fr) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112559689A (zh) * | 2020-12-21 | 2021-03-26 | 广州橙行智动汽车科技有限公司 | 一种基于车载问答的数据处理方法和装置 |
CN112926326A (zh) * | 2021-02-20 | 2021-06-08 | 深圳追一科技有限公司 | 命名实体识别方法、装置、计算机设备和存储介质 |
CN112988993A (zh) * | 2021-02-19 | 2021-06-18 | 车智互联(北京)科技有限公司 | 一种问答方法和计算设备 |
CN113239188A (zh) * | 2021-04-21 | 2021-08-10 | 上海快确信息科技有限公司 | 一种一套金融交易对话信息分析技术方案 |
CN114077831A (zh) * | 2020-08-21 | 2022-02-22 | 北京金山数字娱乐科技有限公司 | 一种问题文本分析模型的训练方法及装置 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050050469A1 (en) * | 2001-12-27 | 2005-03-03 | Kiyotaka Uchimoto | Text generating method and text generator |
US20070192309A1 (en) * | 2005-10-12 | 2007-08-16 | Gordon Fischer | Method and system for identifying sentence boundaries |
US20140172417A1 (en) * | 2012-12-16 | 2014-06-19 | Cloud 9, Llc | Vital text analytics system for the enhancement of requirements engineering documents and other documents |
US20150340033A1 (en) * | 2014-05-20 | 2015-11-26 | Amazon Technologies, Inc. | Context interpretation in natural language processing using previous dialog acts |
US20180314689A1 (en) * | 2015-12-22 | 2018-11-01 | Sri International | Multi-lingual virtual personal assistant |
-
2019
- 2019-12-23 WO PCT/US2019/068421 patent/WO2020139865A1/fr active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050050469A1 (en) * | 2001-12-27 | 2005-03-03 | Kiyotaka Uchimoto | Text generating method and text generator |
US20070192309A1 (en) * | 2005-10-12 | 2007-08-16 | Gordon Fischer | Method and system for identifying sentence boundaries |
US20140172417A1 (en) * | 2012-12-16 | 2014-06-19 | Cloud 9, Llc | Vital text analytics system for the enhancement of requirements engineering documents and other documents |
US20150340033A1 (en) * | 2014-05-20 | 2015-11-26 | Amazon Technologies, Inc. | Context interpretation in natural language processing using previous dialog acts |
US20180314689A1 (en) * | 2015-12-22 | 2018-11-01 | Sri International | Multi-lingual virtual personal assistant |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114077831A (zh) * | 2020-08-21 | 2022-02-22 | 北京金山数字娱乐科技有限公司 | 一种问题文本分析模型的训练方法及装置 |
CN112559689A (zh) * | 2020-12-21 | 2021-03-26 | 广州橙行智动汽车科技有限公司 | 一种基于车载问答的数据处理方法和装置 |
CN112988993A (zh) * | 2021-02-19 | 2021-06-18 | 车智互联(北京)科技有限公司 | 一种问答方法和计算设备 |
CN112988993B (zh) * | 2021-02-19 | 2023-10-20 | 车智互联(北京)科技有限公司 | 一种问答方法和计算设备 |
CN112926326A (zh) * | 2021-02-20 | 2021-06-08 | 深圳追一科技有限公司 | 命名实体识别方法、装置、计算机设备和存储介质 |
CN112926326B (zh) * | 2021-02-20 | 2024-01-19 | 深圳追一科技有限公司 | 命名实体识别方法、装置、计算机设备和存储介质 |
CN113239188A (zh) * | 2021-04-21 | 2021-08-10 | 上海快确信息科技有限公司 | 一种一套金融交易对话信息分析技术方案 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20200143115A1 (en) | Systems and methods for improved automated conversations | |
US11010555B2 (en) | Systems and methods for automated question response | |
US11663409B2 (en) | Systems and methods for training machine learning models using active learning | |
US20210201144A1 (en) | Systems and methods for artificial intelligence enhancements in automated conversations | |
US20200143247A1 (en) | Systems and methods for improved automated conversations with intent and action response generation | |
US11574026B2 (en) | Analytics-driven recommendation engine | |
US20190180196A1 (en) | Systems and methods for generating and updating machine hybrid deep learning models | |
US20200143265A1 (en) | Systems and methods for automated conversations with feedback systems, tuning and context driven training | |
US20190286711A1 (en) | Systems and methods for message building for machine learning conversations | |
US20190272269A1 (en) | Method and system of classification in a natural language user interface | |
US10360305B2 (en) | Performing linguistic analysis by scoring syntactic graphs | |
US11507850B2 (en) | System and method for call centre management | |
US20190179903A1 (en) | Systems and methods for multi language automated action response | |
WO2019113122A1 (fr) | Systèmes et procédés d'apprentissage automatique amélioré pour des conversations | |
US9575936B2 (en) | Word cloud display | |
US9196245B2 (en) | Semantic graphs and conversational agents | |
US10803479B2 (en) | Systems and methods for management of automated dynamic messaging | |
US20190286712A1 (en) | Systems and methods for phrase selection for machine learning conversations | |
US10692016B2 (en) | Classifying unstructured computer text for complaint-specific interactions using rules-based and machine learning modeling | |
WO2020139865A1 (fr) | Systèmes et procédés pour des conversations automatisées améliorées | |
US20200272791A1 (en) | Systems and methods for automated conversations with a transactional assistant | |
US11551188B2 (en) | Systems and methods for improved automated conversations with attendant actions | |
US20190286713A1 (en) | Systems and methods for enhanced natural language processing for machine learning conversations | |
US20200201913A1 (en) | Systems and methods for improved automated conversations with roi metrics and threshold analysis | |
US11106871B2 (en) | Systems and methods for configurable messaging response-action engine |
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: 19906537 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: 19906537 Country of ref document: EP Kind code of ref document: A1 |