Project Kickoff Meeting
The Project Kickoff Meeting introduces the project and describes the project goals. Project Sponsors organize and lead these meetings.
Working collaboratively, the team discusses business issues, linguistic terms, and data sources which contribute to supporting the project’s goals.
As the discussion unfolds, the team develops a “Mind Map”, recording the details of the project and the types of sources necessary to support it. These Mind Maps provide direction for future activities.
When the meeting concludes, each team member will leave with action items for moving the project forward.
Action Items
Subject Matter Experts are asked to locate and compile text examples to capture. Text examples include terms, phrases, and paragraphs of what text to model and what to exclude based on context.
IT Resources are asked to locate data sources and determine how to acquire the data tables produced from the source. If they are unable to acquire data tables directly, they make note of the data source and ask the General IT Resources for assistance.
General IT Resources are asked to make contacts with owners of internal data sources to obtain access to the data tables. Additionally, they will evaluate areas within the IT infrastructure where the model code can be deployed. They may set up data structures for capturing model output, once the models are placed into production.
Linguists, if included on the team, are asked to create code rule taxonomies. Subject Matter Experts or Dedicated IT Resources assume this duty if Linguists aren’t included.
Team Member Activities
General IT Resources
General IT Resources investigate data sources, owners, and methods for acquiring various data tables. Once the data is found and access is provided, General IT Resources will complete extract, transform and load actions on the data to move it into a new data table. Several sources may be combined and stored within a single table.
If time and skills allow, the data within the combined table data may be cleaned and standardized. These can be time consuming activities, but they ensure that high quality data feeds modeling activities; garbage in, garbage out. Once the data table is ready, Dedicated IT Resources are provided access to the table.
Subject Matter Experts
Subject Matter Experts comb through various document sources searching for text examples to model. Specific terms, phrases, or text combinations are highlighted to show examples that the models should include and exclude. These examples are stored as screenshots, or as marked up paper copies.
These are collected and sent to Dedicated IT Resources in a bulk delivery, or as small batches delivered periodically. Well-organized Subject Matter Experts may categorize these examples within a spreadsheet for tracking as the model is developed.
Dedicated IT Resources
Once the data table and document examples are received, the Dedicated IT Resource becomes the primary user of the application. They open the application, start a new project, and connect the data table. Now they can begin investigating text contents and begin writing model code.
Savvy IT Resources will exit the application at this point and engage Linguists and Subject Matter Experts to plan the taxonomy contents before writing model code. Well-planned taxonomies save trouble later in the process.
Inexperienced IT Resources, however, will plunge forward into the project, writing model code and building taxonomies in an ad hoc fashion. They continue building as they work to create code entries which include and exclude examples based on context. Those who muddle forward experience significant trouble later in the process. Most choose this path, and trundle forward.
The coding process proceeds, and a model is ultimately created. The model is an abstract code block at this point, a work-in-process material; no business value is produced at this stage.
Dedicated IT Resources send the model code to General IT Resources, who place it into production.
Model In Production
The activities in this portion of the process are much less clear, as the model becomes a “black box” ingesting data and outputting a categorized data table.
Categorized data tables are the value producing component of this operation. Value extraction depends on the business goals of the organization.
Once a sufficient table is produced, it may be sent to Dedicated IT Resources for cleaning and deduplication. The amount of time and effort spent in this portion of the task is dependent on the amount of time spent in planning. More time spent in planning taxonomies, less time and effort spent in post-production cleaning and deduplication activities.
Model Management and Maintenance
The activities in this portion of the process are much less clear, as the model becomes a “black box” ingesting data and outputting a categorized data table.
Categorized data tables are the value producing component of this operation. Value extraction depends on the business goals of the organization.
Once a sufficient table is produced, it may be sent to Dedicated IT Resources for cleaning and deduplication. The amount of time and effort spent in this portion of the task is dependent on the amount of time spent in planning. More time spent in planning taxonomies, less time and effort spent in post-production cleaning and deduplication activities.
Model Retirement
End of life activities for the model are not notable. The model is simply removed production and discarded or replaced with something more effective. This step was also ignored in the process.
Related Examples
User Profiles
This example discusses what the product does, it’s history, and how it provides value within an organization. This information provides the reader with a background for understanding the other examples within this section.
Architecture Redesign
Development conceived this product being used by a single person. Research, however, revealed something different. This story describes the collaboration process companies use to extract value from this product.