Skip to content

Message topic classification agent

This is an agent that performs the message classification by topic (greeting, about lab work condition, about entity etc.). Also, the agent classifies message by trait (neutral, positive, negative emotional coloring) and gets message entities. The agent classifies a message using Wit.ai.

Action class:

action_message_topic_classification

Parameters:

  1. messageAddr -- an element of concept_message.

Libraries used:

  • Wit.ai - to classify message and get entities.

Comment:

  • The input message must contain a text file with the text in Russian;
  • The excluding entity should be formalized in knowledge base.

Examples

Examples of an input structure:

Examples of an output structure:

Example of a structure needed to classify message by intent:

Example of a structure needed to classify message by trait:

Example of a structure needed to get message entity:

Message classes:

  1. concept_greeting_message
  2. concept_message_about_entity
  3. concept_message_about_subdividing
  4. concept_message_about_study_system
  5. concept_message_about_lab_work_condition
  6. concept_message_about_lab_work_deadline

Agent implementation language

C++

Result

Possible result codes:

  • SC_RESULT_OK - the message is successfully classified (or there is empty classification) or the action doesn't belong to the action_message_topic_classification.
  • SC_RESULT_ERROR- internal error.