Dcipher Analytics is a comprehensive toolbox designed to automate research and insight work. It contains a variety of tools, each designed for effective analysis of large text sets. Some of these tools, known as Dcipher Workflows, are specialized for niche applications, covering tasks that range from compiling summaries of global news to analyzing open-ended survey responses. The workflows can generate output in different formats, from user-friendly chatbots to dense tables packed with information. In addition to the workflows, the toolbox boasts the Dcipher Studio, an exceptionally versatile tool capable of handling essentially any type of text analysis.
The diagram shown above illustrates the key dimensions you need to consider when determining which Dcipher workflow is the most appropriate for your specific use case. This article guides you through the process.
Dcipher Workflows or Dcipher Studio?
As a general rule, Dcipher Workflows are recommended when a specific workflow exists that fits your needs. However, in some cases you may want to train your own AI models and build your own customized text analytics pipeline for a specific task — this is when you will need the full capability and flexibility of Dcipher Studio. For more information about Dcipher Studio, visit our comprehensive Studio Help Center guide. To understand how to select the right Dcipher Workflow for your needs, keep reading this article.
What is the overarching focal question?
To pick the right insight automation workflow in the Dcipher toolbox, we need first to be clear about our overarching focal question. Do we want to…
…map conversations around a particular topic or brand?
…scan the horizon for trends and weak signals?
…understand the voice of customers?
…make knowledge within or or outside our organization more accessible?
These are just a few of the insight tasks that can be automated using Dcipher Workflows.
What information sources are useful for answering the focal question?
Based on the focal question, what information sources are most relevant? Do we want to use…
…news to map out reporting around topics and brands, or to identify trends and weak signals?
…public social media to understand online conversations?
…documents or reports (in Word or pdf format) to gain insights from information stored in our organization’s digital archive or from external sources?
…datasets (in Excel, csv, tsv, or json format) with unstructured text data, such as free-form text responses in surveys?
Once you've determined the information source, you can narrow down the Dcipher tools that apply to your specific source. If you are interested in multiple source types, it is recommended to make separate runs — one for each source type — to make the results as interpretable as possible.
What approach is most suitable – bottom-up or top-down?
Are you aiming to uncover patterns and detect “unknown unknowns” within a broad field, with no predefined questions or topics of interest? If so, a bottom-up analytical approach is most suitable for your needs.
Or are you, on the contrary, more interested in “finding needles in the haystack,” for example getting answers to specific questions or getting summaries of content about well-defined topics or brands? If so, a top-down analytical approach will serve you better.
Is the analysis one-off or recurrent?
Consider whether you need a one-time analysis, or if you would like the analysis to be automatically re-run, say, every day or every week? While running a workflow a single time is a convenient and efficient solution to many research tasks, setting up a recurrent analysis that runs automatically at your chosen intervals can be just as straightforward — and some workflows are designed specifically for that. Recurrent analysis is especially useful when working with data sources that constantly update, such as newspaper articles or social media posts.
Examples:
1. You want to map conversations around your brand in public social media
Source: Social media
Approach: Bottom-up
Recurrency: One-time
2. You want to identify themes in free-form text responses in a survey
Source: Dataset
Approach: Bottom-up
Recurrency: One-time
Workflow: Free-form text survey response analysis
3. You want to identify trending topics in the news reporting in an area of interest on a daily basis
Source: News
Approach: Bottom-up
Recurrency: Daily
Workflow: Daily news updates
4. You want to get a monthly summary of a set of predefined topics of interest in the news
Source: News
Approach: Top-down
Recurrency: Monthly
Workflow: Topic analysis based on news reporting (scheduled runs mode)
5. You want to get answers to a number of specific questions in relation to a topic of interest based on close to a hundred PDF reports
Source: PDF reports
Approach: Top-down
Recurrency: One-time
Workflow: Research bot based on PDF reports