Articles in this section

DataLens for AI Powered Text Search

Avatar
Mark Dyer
  • Updated
Follow

DataLens enhances search functionality by leveraging AI to surface the most relevant records, making it a powerful tool for identifying trends and patterns in large datasets. DataLens is an AI-powered text search tool designed to return results based on approximate matches to a given theme. Unlike traditional search methods, DataLens analyses and ranks records based on their relevance, allowing users to retrieve the most contextually appropriate results. By enabling DataLens, users can move beyond keyword-based searches and gain deeper insights into their data.

 

To perform the search click on the AI icon in the report search bar. This switches the search method to be DataLens as opposed to the standard Quick Search Report feature. This action enables AI-driven search, allowing broader and more relevant results beyond exact keyword matches.
 

Use Case: Identifying Thematic Issues

Here we apply this to the use case of identifying Issues that relate to a theme. Our App contains hundreds of Issues which have been uploaded and modelled so that when we now search our data it will know which Records relate to our search term.

 

In this example we are looking for all issues that relate to the theme of 'Quality Control'. If we were performing this via the standard Quick Search capability then this would only bring back the results were the text entered is an exact case insensitive match to text that appears in the Fields that we have selected to be in the Quick Search subset of Fields. By using the DataLens search capability, the Records returned will include the top 40 records in order of relevance to quality control.

 

DataLens optimizes the report view by sorting the results based on relevance. It truncates non-matching records to highlight the most relevant ones. It displays a horizontal bar beneath each record’s action button, where a longer bar indicates a closer match to the search term and a shorter bar represents a weaker match.

 

 

Configuration

DataLens is enabled from the 'Advanced Config' tab within the 'App Configuration' Section of the App build.

 

Enabling DataLens

Enable DataLens: Toggling this option on will present the User with the ability to switch the Report into the AI DataLens search capability

Enable DataLens Translation: Many applications will have Users in global regions. DataLens is capable of handling any language to a decent degree and enabling this option makes it easier for the User as they are able to search for Records in their native language

 

DataLens Fields 

DataLens will take the values from all the Fields that you define here and model the DataLens to search based on this so that when the search is performed it brings back approximate matches to the data across all of these Fields. 

 

Tip: For best results, use fields that contain plenty of text (e.g. one or more sentences), and search using whole sentences rather than one or two words

Limitation: Currently, if you use App Studio to add a new field to DataLens in an app that already has DataLens enabled, existing records' values for that field will not be indexed until they change. Therefore, be sure about your choice of fields before doing the app publish that enables DataLens. If you have already enabled DataLens and need to add a new Field then the best way to include the data in the DataLens search will be via export and import changing the value in the additional Field away from and then back to the correct values meaning the system will recognise a change in value and update the DataLens model.

 

 
Considerations
The DataLens search index is updated asynchronously. For changes made one record at a time e.g. editing in Workspace, the index should be updated within a minute or two of making the edit. For large batches of changes, re-indexing is not immediate but rather is done in a scheduled basis our servers run once an hour.
 
DataLens index generation is further rate-limited due to the costs and limits of the AI service that powers it. A very large change (e.g. importing several hundred thousand records) could take many hours to be fully indexed.
 
DataLens, in its current state, does not recognise numbers well and searching DataLens for a number value will not give sensible results. Say we wanted to use DataLens in our Issues App to identify all Records where the value of the impact is a 3. By searching for 'Impact - 3' it is unlikely to return the correct results. For this type of analysis it is better to use the traditional Report search capabilities. It is worth noting this is a general limitation of the current state of AI searches as a whole – we anticipate this should improve over time.

Related articles

Comments

0 comments

Please sign in to leave a comment.