The Sensis Ontology Program is an enabler for richer classification of Business, Government and Geospatial content and data across Sensis which aims to move classification away from manually managed printed books into automated digital clusters of terms, concepts and categories. The Ontology Program is designed to assist in providing fresher, more accurate and deeper content (e.g., Brands and Products) for listing retrieval via the Sensis Search API, as well as supporting data for business intelligence and data science teams.
As part of the Sensis Ontology project, I have been in charge of the following tasks:
exploring tools, standards and services to build a commercial Ontology;
planning and monitoring development and testing at project-level to implement the production-level Ontology;
integrating the Ontology with other repositories and capabilities, including the Sensis Search API, to enhance internal products and processes;
devising and interacting with business units within Sensis for new products based on Ontology data.
Technical implementation of the Ontology covers tasks in Natural Language Processing and Ontology construction, to name some:
Terminology Extraction and Analysis:
devising algorithms for retrieving meaningful content from unstructured texts;
using part-of-speech analysis, term weighting, and reference data to extract relevant keywords;
Concept Detection and Clustering:
- naturally discovering and expanding clusters of terms that represent similar concepts.
Some of my accomplishments in this project have been:
proposing a first prototype of the Ontology in six months, and a full-fledged production system in a year, which was overall half of the time expected at the start of the project;
using an early version of the Ontology in a prototype for a company-wide innovation event where our team got awarded the first place in the “Innovation for Customers” category;
implementing several prototypes based on internal customers’ requirements, such as:
data visualisation and process flow tools to identify overlaps and gains between pre-existing content and Ontology concepts;
database systems that contained specialised information that represented brands, services and products acquired with the Ontology;
APIs to access information in the Ontology for internal consumption.