Introduction

Whisper (Qingyu) is a cloud service platform based on natural language understanding (NLU) and artificial intelligence (AI) dialogue technology. Whisper provides the means for enterprises to conveniently build chatbots. Even without deep knowledge of NLU and AI techniques, users can create and publish their own chatbots to websites, mobile applications, smart devices, etc. With the help of Whisper.

Benefits

Understanding

  • Industry-leading semantic similarity analysis technology applied in problem finding, clustering, generalization of similar sentences, etc.
  • With our unique sentence analysis algorithm, the platform accurately supports entity extraction and intent classification based on simple grammar templates. Combined with our semantic similarity algorithm, flexible customizations can be achieved.
  • Deep neural network model supports document comprehension, self-learning and continuously optimization

Conversation

  • Flexible skill combinations
  • Supports contextual understanding and multiple rounds of conversations
  • Supports entity slot operations in template answers
  • Supports third-party answer services
  • Supports rich components (multimedia outputs)

Service

  • Based on containerized and clustered deployment, the platform is easily maintainable and scalable
  • High availability, reliability, performance
  • Support privatized deployment
  • All-in-one working process: build, train, test, release within Whisper platform
  • Quick connection with multiple interfaces

Basic Concepts

Bot: Corresponds to an independent and complete dialogue system with independent personality attributes configuration. Multiple skills can be selected.

Skill: Skills are a set of question and answer pairs or a set of intentions to accomplish a specific task in a specific scenario.

Paired Corpus: A paired corpus consists of a set of similar questions and a set of answers. A paired corpus can be labeled with multiple classification labels.

Intent: Purpose that the user wants to express in the dialogue. The service organizes intents by the labelled slots in the question and answer templates.

Slot: The key information or restrictive conditions contained in the user's expression of intents. The system is represented as an entity dictionary. Users can customize the dictionary or use the built-in dictionary.

Organization of dialogue within the platform

  • Each user can create multiple bots
  • Each user can create multiple skills (In current version, two default skills are provided)
  • Each chatbot can be configured with multiple skills
  • Each skill can contain multiple intents or multiple Q&A pairs
  • Customized synonyms and entity dictionaries can be set for each skill
  • Multiple slots can be labelled within each intent