Unlimited Reporting with Natural Language Processing (NLP)

The number of reports generated by an HRMS solution is one of the important criteria in buying or adopting an HRMS solution. There are 100s of HRMS solutions in market and the number of reports offered by them varies from 20 to even 200 reports. While having more reports is a plus, no solution can offer everything that an HR Manager or management needs for their custom and bespoke reporting requirements.

To overcome this problem, Runtime HRMS came up with a novel idea - unlimited reports! Yes, you read it right. Runtime HRMS can generate unlimited number of reports based on user requirement. We utilize Machine Learning (ML) and Natural Language Processing (NLP) to generate any type of report a user may need. What's more exciting about this feature is that you do not need technical knowledge to generate custom report formats, select reporting columns and add filters. All you need to do is - ask your question. Our NLP engine understands your query and generates reports dynamically on-the-fly.

Terminology

Take a quick look at various terms used in this article for better understanding:

1. NLP - Natural Language Processing
2. LLM - Large Language Models
3. Query Resolver Middleware (QRM): Our middleware that understands user questions asked in natural language and converts them into database queries.

Natural Language Processing (NLP)

NLP stands for Natural Language Processing, which is a branch of Artificial Intelligence that allows computers to understand, generate, and manipulate human language. The technology can accurately extract information and insights contained in the documents as well as categorize and organize the documents themselves.

How do we use NLP for reporting in Runtime HRMS?

We make use of Large Language Models (LLM) to understand user queries and convert them into meaningful database queries. For e.g. if a user asks for the list of employees with O+ve blood group, our Query Resolver Middleware (QRM) generates and thoroughly validates the SQL statement to run against the database. We then fetch results from database and perform some quick security checks to ensure data integrity. Once the results are filtered/scanned, the data is presented in a tabular format by Runtime HRMS. All results generated using this feature can be downloaded as an Excel file with just a click.

Practical use cases of NLP reporting

Let's dive down into some amazing use cases of NLP reporting with Runtime HRMS. 

Let's say you want to see the diversity ratio among employees. 
Start by asking: What is the gender diversity?

Here are the results:

Now, let's say you want to know how many people took CL in previous month. 
Ask: How many people took CL last month?

The important thing to note here is that when you specify 'last month', the QRM automatically figures out the month based on current date. If you want to query for a specific month, mention month and year in your query. For e.g. How many people took CL in Jan 2024? 

Let's take another interesting example. Here, we will test the model's intelligence on understanding manager-supervisor relationships. You can ask: How many people report to Manmeet? 
(You should use actual manager name instead of Manmeet)

If you want list of people reporting to Manmeet, change your query to: List of people reporting to Manmeet

You will notice that Manmeet could be first name, middle name or last name. But our QRM understands the query intent and finds out people reporting to manager named Manmeet. To be more specific, you can mention employee code also. For e.g. List of people reporting to employee code EMP001.

Let's try something more complicated by asking this:
Mumbai employees who were absent for more than 5 days in last month. Also include no. of days they were absent.
(Replace name of location as per your data)

You see, the possibilities are unlimited with custom reporting using NLP. We welcome you to use your imagination and give it a try today.

Way Forward

Every report generated by NLP will ask the users for their feedback. Its simply a up-vote or down-vote. This helps us understand how performant the reporting is and what are the areas that need improvement. The feedback collected from users is then used by the system to further improve results over time. The system is designed in such a way that the most up-voted queries over time will become preferred answers to other similar queries. Thus, the system is self-evolving.

This is not something that we built once, fire and forget. This feature uses technologies that are on the verge of bleeding edge and are constantly evolving. We commit to continuously monitor, improve and experiment to make the most of it in times to come.