How to build an NLP use case?
When you log into the Accern Platform, you will see the Dashboard screen with a Welcome video and status of current use cases and issues.
Click on the Deploy tab to view all use cases or start a new one.
Create New Use Case
Click on the Build New Use Case blue tile to start a new use case. Select the Adaptive NLP option to classify, extract and derive analytics from unstructured data. Define a use case name and category, then click Get Started to proceed.
You can also customize your own Use Case category:
The first step of building the use case is selecting one or multiple data sources that you would like to extract insights from. Toggle on/off the switch button on each tile to make your selection. By default, we have four out-of-box tiles: Public News, Public Blogs, NewsEdge, and DowJones. SEC Filings are available upon request.
Foreign Language: is available for Public News and Blogs.
Hover over the 3 dots and select Configure. Choose the language(s) and click Save & Apply.
After selecting the data sources, you will need to select the taxonomies to determine what to extract out of the unstructured data. There are two types of taxonomies: Entity (US Equity, International Equity, Cryptocurrency, Forex, and Commodity) and Themes/Events.
There are multiple ways to select taxonomies.
1. Using Search Bar: type the entity name onto the search bar and check the box to select
2. Select by asset classifications: entity taxonomies are also grouped into asset classifications, indices, sectors, industries, and exchanges. You can select the entire group of taxonomies from the list we have out-of-box.
Themes taxonomies are divided into Event Groups. We have 38 financial services industries' focus event groups available out-of-box. You can select the entire group or individual sub-events under each group.
3. Bulk Upload and Taxonomy Customization: visit this documentation
After selecting the taxonomies, proceed to the NLP Models page where all the pre-trained models reside. For best results, we highly recommend turning on all model packages (except Foreign Language Translation unless you configure your data sources with a foreign language). You can also select specific features under each package as well by simply checking/unchecking the boxes.
To finish creating a use case, you need to configure the delivery methods.
First, select use case type:
Real-Time: a streaming pipeline that will continuously generate data from when you start the use case
Historic: process the data sources in a specific time range. When you select this option, a window will appear for you to define your time range for each data source selected. Click Close to save.
Second, select the use case output destination. There are many options available, including API feed, databases, file stores, and BI tools.
We also offer the Accern Data Store option where the data will be exported into an API feed and a dashboard. To generate an Accern Data Store, simply click on “Use Accern Data Store” and the platform will automatically assign a name to the Delivery Tile that matches the use case name. Click Add to save your selection.
To select/deselect the delivery selection, toggle on/off the button on the Delivery Tile. You can select multiple delivery destinations at once for each use case.
For other delivery methods using databases, file store, please visit the documentation below:
Review Use Case
After completing the configuration for your use case, click on Review Use Case on the top right corner.
A window will pop up for you to review your selection. Click through each tab (Data Store, Taxonomy, NLP Models, Delivery Integrations) for more details. Click Run Use Case to proceed. If your use case successfully starts, the Use Case Data window will appear.