Abstract:
Disclosed are systems, methods, and non-transitory computer-readable storage media for utilizing a virtual assistant as part of a communication session. One or more of the participant users can select to utilize a virtual assistant to assist the participant users with tasks during the communication session. A user can use a communication application to enter a message directed to the virtual assistant. The virtual assistant can analyze the entered message and determine that the message was directed to the virtual assistant rather than to the other participants of the communication session. As a result, the message will not be transmitted to the other participants of the communication session and the virtual assistant will assist the user with the identified task. A virtual assistant can assist a user with a variety of different tasks.
Abstract:
A dialog manager receives previous user actions and previous observations and current observations. Previous and current user states, previous user actions, current user actions, future system actions, and future observations are hypothesized. The user states, the user actions, and the user observations are hidden. A feature vector is extracted based on the user states, the system actions, the user actions, and the observations. An expected reward of each current action is based on a log-linear model using the feature vectors. Then, the current action that has an optimal expected reward is outputted.
Abstract:
An approach is provided for mining threaded online discussions. In the approach, performed by an information handling system, a natural language processing (NLP) analysis is performed on threaded discussions pertaining to a given topic. The analysis is performed across multiple web sites with each of the web sites including one or more threaded discussions. The analysis results in harvested discussions pertaining to the topic. The harvested discussions are correlated and a question is identified from the harvested discussions. A set of candidate answers is also identified from the harvested discussions, with one of the candidate answers being selected as the most likely answer to the identified question.
Abstract:
In language evaluation systems, user expressions are often evaluated by speech recognizers and language parsers, and among several possible translations, a highest-probability translation is selected and added to a dialogue sequence. However, such systems may exhibit inadequacies by discarding alternative translations that may initially exhibit a lower probability, but that may have a higher probability when evaluated in the full context of the dialogue, including subsequent expressions. Presented herein are techniques for communicating with a user by formulating a dialogue hypothesis set identifying hypothesis probabilities for a set of dialogue hypotheses, using generative and/or discriminative models, and repeatedly re-ranks the dialogue hypotheses based on subsequent expressions. Additionally, knowledge sources may inform a model-based with a pre-knowledge fetch that facilitates pruning of the hypothesis search space at an early stage, thereby enhancing the accuracy of language parsing while also reducing the latency of the expression evaluation and economizing computing resources.
Abstract:
A scalable statistical language understanding (SLU) system uses a fixed number of understanding models that scale across domains and intents (i.e. single vs. multiple intents per utterance). For each domain added to the SLU system, the fixed number of existing models is updated to reflect the newly added domain. Information that is already included in the existing models and the corresponding training data may be re-used. The fixed models may include a domain detector model, an intent action detector model, an intent object detector model and a slot/entity tagging model. A domain detector identifies different domains identified within an utterance. All/portion of the detected domains are used to determine associated intent actions. For each determined intent action, one or more intent objects are identified. Slot/entity tagging is performed using the determined domains, intent actions, and intent object detector.
Abstract:
The present invention relates to a system and method for linguistic processing. The system (100) comprises a representation processor (110) having an abstractor component (120) and a specializer component (130), a stored mappings database (140), a standard vocabularies database (150), and a linguistic ontologies database (160). The abstractor component (120) transforms a linguistic representation into an abstract representation by using abstraction rules. The abstractor component (120) includes a concepts and properties extractor component (121), a verb determinator component (122), a schemas extractor component (123), a schemas mapper component (124), and a concepts and properties matcher component (125). The specializer component (120) transforms an abstract representation into a linguistic representation by using specialization rules. The specializer component (130) includes a triple extractor component (131), a property matcher component (132), a verb determinator component (133), a verb mapper component (134), a semantic roles mapper component (135) and a triple assembler component (136).
Abstract:
본 발명은 미리 지정된 특정 목적이나 서비스를 수행하기 위한 것이 아닌 자유 발화에 대해 인공지능 에이전트와 대화가 가능한 채팅(chatting) 대화 시스템을 구축하는 데 있어서 효과적으로 대화 관리를 수행하기 위한 모델 구축 방법에 관한 것으로서, 기존의 대화 관리 시스템을 구축하기 위해 훈련용 말뭉치에 대해 필요했던 사람의 수동 태깅 (DA annotation, lexical pattern 규칙 등)을 줄일 수 있는 새로운 방법, 채팅 대화의 특성에 특화된 렉시컬구문패턴(LSP; Lexico-Syntactic Pattern) 생성을 이용한 발화 표현 방법, 개체명 (Named Entity, NE) 관련 태깅 방법 및 정책, 시스템의 인격을 나타내는 persona 적용 응답 생성 방법, 이를 이용하여 실제 대화 관리에서 적절한 시스템 발화를 생성할 수 있는 대화 관리 모델을 제공하는 데 그 목적이 있다.
Abstract:
A method and system for teaching a user a target language includes developing and constructing variable potential paths of nodes representing an exchange between two participants in a dialogue, prompting and selecting a path of nodes through a conversation graph of the target language, the path of nodes defining a dialogue; and determining whether the user is ready to perform the dialogue that has been constructed and defined by the path of nodes, the determination being based on a user model which represents the user's current ability in and current knowledge of, the target language. If the user is ready to perform the dialogue, the path of nodes is executed to allow the user to perform the dialogue defined thereby; and if the user is not ready to perform the dialogue, training the user on one or more nodes of the path of nodes.