Sentiment Analytics Pipeline to help Not-for-Profit Organisations and People

Context

Australian Not-for-Profit organisations employ 1.3 million people and 3 million volunteers to help the homeless, educate the children, feed the hungry, support the arts, enable sports and community, among other. To help them become better at what they do, a UWA initiative called `Learning for Purpose' developed and deployed a first generation suite of People Analytics. Using a scaleable web-based survey and reporting system, we provide - for free - Not-for-Profit organisations the opportunity to measure, diagnose and change the states of their worker well-being, learning, and happiness. See learningforpurpose.org/analytics.

Problem

People do not always voluntarily or timely communicate concerns or problems. The unsystematic collection of and dealing with such (hidden) emotional information can deprive organizations of the potential to successfully address situational and motivational issues which in turn can fuel disengagement, sabotage, or attrition. The traditional measurement of worker sentiments is based on quantitative questionnaires that carry methodological disadvantages (e.g., low fidelity).

Approach

Technology can augment the measurement of affective signals through the computerized analysis of natural language contained in, for instance, free text responses in surveys, or even emails, chats, and transcripts of interactions. Existing named entity extraction algorithms can isolate meaningful phrases from natural language text that correspond to people, objects, places, and events (e.g., boss, cafeteria, John, pay rise, meeting length). Sentiment analysis can determine the affective polarity (i.e., positive, neutral, negative) and magnitude (e.g., 0 to 1) transmitted alongside these entities.

Goal

The project team is asked to build an end-to-end analytics pipeline for sentiment analysis using various existing technologies (e.g., Google Cloud Natural Language API). The intellectual and engineering contribution is to:
  1. understand different user needs: researcher seeking raw scores vs. manager seeking applied insights
  2. stitch together a series of cloud-based systems: Qualtrics survey system, Google BigQuery, web page/sheets
  3. test and iterate with applied data, visualise online, document to enable future development
Likely involves existing and bespoke code in JS, HTML5, CSS3, node.js

Benefit

The advanced analytics of employee and volunteer sentiments enables Not-for-Profit organisations to better understand the phenomena being discussed by staff in the natural environment, the tone, and change over time, including detection of potential issues and opportunities for maintaining and retaining the workforce, and thus helps them do more good in this world. Such a system may also be used to enhance student feedback and learning, among other. Dr Ramon Wenzel operates at the nexus of people science, analytics, and tech. A functional end-to-end implementation into the existing Workforce Analytics Dashboard at Learning for Purpose is rewarded with $500 for the project team.

Client

Contact Person: Dr Ramon Wenzel (Research Ass/Professor, Business School, Centre for Social Impact)
Telephone: 08 6488 5675
Email: [email protected]
Preferred method of contact: any
Location: on campus 39 Fairway, Crawley | Tue-Fri

Client Unavailability

None

IP Exploitation Model

The client wishes to use a Creative Commons CC BY-NC model to deal with IP embodied in the project.