MonkeyLearn's Workflow Builder allows users to build their own text analysis workflow to enrich and analyze their text. These are the high level steps you'll want to follow to build your own workflow within the MonkeyLearn app.
Upload your training dataset from a CSV file.
Clean your data of noise, like a greeting or sign-off, or PII, like a name or email
Split your data into smaller snippets of text.
Enrich your data with models, there are different models you can use to enrich your data, including sentiment, topics, keywords and extraction.
Train your models, you'll want to teach the models how to read your unique text, so the model can learn from your input.
Process/reprocess your data: once you have trained a model, you'll want to process your data (or reprocess if you've made changes to a model).
Visualize your data: your data will feed into a dashboard for analysis, where you'll be able to customize the look and feel of the dashboard's charts and tables.
What type of data can I upload?
You can upload any text data. The most common data types we see are open-ended text responses, like survey responses or support tickets. We also see open-ended text from reviews, NPS (Net Promoter Score), or CSAT (Customer Satisfaction) that can be analyzed alongside a numeric rating.
Why would I need to clean my data?
Oftentimes, your text might include words, phrases, or sentences that are considered "noise" that can generate noise and confusion to a model. Things like, "Hello,", "This is an automated response", or "Submitted at 4:29pm" could all be considered noise and can be easily removed using filters. You may also want to remove PII (Personally Identifiable Information) to keep any names, emails, or document names out of your analysis.
What does splitting my data mean?
You may want to split your data into smaller snippets of text. If one piece of text contains multiple different opinions or sentiments, or if your text can be categorized as containing multiple topics (for example, a review that says, "The coffee was great but the service was terrible.", you might want to split it into two different opinions, the first talking positively about the coffee and the second talking negatively about the service.
Which models should I use to enrich my data?
This depends on what type of analysis and insights you'd like to get from you data, and what type of data you're importing. The most commonly used models are ones that analyze sentiment (whether the text skews positive, neutral, or negative), keywords (what keywords are most common in your text), and topics (what category or topic is mentioned the most in your text).
How do I "train" a model?
A lot of the models we provide are pre-trained, but models that are specific to your text, like the topic model, need to be trained to understand your unique text in order to tag it with the topics you've specified. You only need to train a small amount of text (this is called your training data) in order for the model to learn from your input and correctly process the remainder of your text.
What happens when my data is "processed"?
Once you've trained the model, you want all of your data to be run through the cleaning filters and models, which is how your data becomes enriched with additional insights and analysis. The models will take what it learned from your training data and apply that to ALL of your data. Any time your train or retrain your models, you'll want to then reprocess your data.
What are my data visualization options?
MonkeyLearn's built-in dashboards provide a variety of different graphs, tables, and charts that let you see trends in your data and let you dig deeper into your enriched data to gather insights you otherwise wouldn't have seen. Our dashboards are customizable in that you can add or remove widgets, change the layout of the dashboard, and filter by any of the metadata fields within your data.
Ready to get started?
Great! To start building a workflow, login at app.monkeylearn.com, navigate to My Workflows, click "+Create Workflow"!