User:Anupama06/sandbox

ONTOLOGY BASED EXTRACTION USING DEEP PARSING IN WATSON

We are developing annotators to identify entities like Symptoms, Action, Advise and Activity from natural language text in support tickets. We need SME help to annotate some of the tickets (as attached) which would serves as the groundtruth for evaluation of the developed annotators.

FOR SME: Description of how to annotate

Example 1: WE HAD A HARD DRIVE FAILURE AND REPLACED IT WIT H A SPARE THAT WE HAD ON HAND. PLEASE SEND A CE ONSITE TOMORROW WITH A REPLACEMENT DRIVE

Symptom: HARD DRIVE FAILURE

Activity: REPLACED IT WITH A SPARE

Action: SEND A CE ONSITE TOMORROW WITH A REPLACEMENT DRIVE

Example 2: Multiple entities of same type. This may require splitting as well.

My disk and fan are not working

Symptom1: Disk not working

Symptom2: Fan not working

Example 3: Trying to install Installation Manager 1.8 in production for CF14

Activity: install Installation Manager 1.8

Example3.png|Highlighted annotation for Example 3

Words Associated to each entity:

Symptom: Fail, not working, cant boot,

Action: send, replace,

Activity: trying, running

Advise: help, advise, let know, how to and all wh-questions