In the last decade, the market of Internet of Things has become quite disruptive, together with commercial clouds providing connection between every sort of devices and the global network, supported by vocal assistants. On the other hands, such commercial products are limited to work on limited domains, although easily scalable, without aspiring to higher level of reasoning in the field of Decisions Making. In this work, we show a way towards the design of an architecture for building cognitive agents leveraging Natural Language Processing. Such agents will be not based on clouds and do not require any semantic training, plus they will be able of deduction on facts and rules in First Order Logic inferred directly from Natural Language. After the description of the architecture and its underlying components, a case-study is provided to show the effectiveness in cases of direct commands and routines, subordinated also by a Meta-Reasoning in a conceptual space, parsing the utterances with promising real-time performances.

在过去的十年中，物联网市场已经变得相当具有破坏性，商业云在语音助手的支持下提供了各种设备和全球网络之间的连接。另一方面，这种商业产品仅限于在有限的领域工作，尽管很容易扩展，但并不渴望在决策领域达到更高水平的推理。在这项工作中，我们展示了一种设计用于构建认知代理的架构的方法利用自然语言处理。这样的智能体将不基于云，不需要任何语义训练，而且它们将能够从自然语言中直接推断出一阶逻辑中的事实和规则。在对架构及其底层组件的描述之后，提供了一个案例研究来展示直接命令和例程的有效性，在概念空间中也由元推理从属，解析具有有希望的实时性能的话语.