Browse Prior Art Database

Method and System for Conversational System Configuration Using Machine Reading and Dialog

IP.com Disclosure Number: IPCOM000251873D
Publication Date: 2017-Dec-07
Document File: 4 page(s) / 456K

Publishing Venue

The IP.com Prior Art Database

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

1

Method and System for Conversational System Configuration Using Machine Reading and Dialog

Abstract

A method and system is disclosed for configuring a system by reading task recipes and schematic user inputs and engaging in a dialog with a user through a conversational interface to execute appropriate commands to complete a task.

Description

Generally, many tasks related to computation systems such as, but not limited to, configuring information technology (IT) asset clusters or installing software application, require users to carry out detailed step by step command line interaction with a system. In many cases, configuring the IT asset with a capability includes well known recipes or recommended procedure for installing and configuring the capability along with considered issues. The recipes are simply list of commands to issue for configuring the system. However, following such recipes is time consuming and is an error prone activity when parameters specific to customer's IT cluster configurations need to be maintained.

Thus, there exists a need for a method and system for generating a document structure by reading task recipes to configure a system that interactively carries out a task through a conversational interface.

Disclosed is a method and system for configuring a system by reading task recipes and schematic user inputs and engaging in a dialog with a user through a conversational interface to execute appropriate commands to complete a task.

The method and system configures a system by using models of document structure, domain knowledge and natural language understanding. In order to configure the system, the method and system generates a logical description of a document/document structure. The logical description of the document needs to be generated in order to understand sequences of actions and choice points. The logical description of the document is generated by reading task recipes that are written in a combination of natural language and code snippets and identifying each block and hierarchical relation to other blocks. The method and system reads the task recipes by using the natural language understanding model.

After generating the logical description of the document, the method and system identifies and labels the sequences of actions and choice points in a workflow described in the document to generate an action model with atomic actions as base elements. The atomic actions are then linked to source document segment. The action model is generated by using analysis of the document and using a domain specific interpretation mechanism that interprets the logical description o...