Table of Contents

Abstract 3

Introduction - 4

Background

Scope

Literature Review - 5

Tools and Technologies - 6

Methodology - 7

Data Preparation

Email Sending Process

Testing and Validation - 8

Implementation 9

Results and Discussion - 11

Conclusion - 12

References - 13
With the recent developments in Large Language Models (LLMs), a lot of work has been enabled across
different fields. As we approach an AI-oriented world, we intend to showcase how the collaboration of
Robotic Process Automation (RPA) and LLMs yields efficient functioning in numerous engagements. This
study combines the Phi3 Language Model in parallel with UiPath Studio to perform web data extraction
followed by enhancing the data to provide more effective and better summaries using the LLM. The
summarized data obtained is then emailed using automated email sending via HTTP requests. This approach
is useful for notification services that need to be repeatedly dispatched across various sources, such as
sourcing and filtering relevant information related to hackathons from different platforms and then presenting
the results in an intuitive and innovative manner. Another application is the comparison of event calendars
across various sites with a single tab that offers comparative information about them in detail.

Index Terms—Large Language Models (LLMs), Robotic Process Automation (RPA), UiPath Studio, Phi3
Language Model, Web Data Extraction, Data Summarization, Automated Email Sending
 Introduction
The rise in participation in technical skill-building workshops, hackathons, coding events, etc., poses a
significant challenge for aspiring professional students who are expected to constantly search for new
activities on various platforms. The problem lies in the fact that such systems are rarely as tightly integrated
as one might anticipate, leading to the possibility of missing important opportunities. This issue can be
addressed by incorporating large language models (LLMs) to automate the discovery and notification of such
events. This requires a solution wherein the user receives a summary of relevant live or upcoming events via
email, thereby simplifying the notification process and ensuring that users are informed about the availability
of opportunities.

Data mining, which is core to this solution, involves retrieving data from websites through web scraping. For
instance, utilizing the features bar in UiPath Studio’s UI Explorer allows data to be pulled and logged into a
usable format, such as a DataTable. This process enables the collection of numerous information sources
related to various events and helps obtain a detailed vision of available opportunities.

One of the advancements in natural language processing is the development of lightweight Large Language
Models (LLMs), also referred to as Small Language Models (SLMs), such as Phi3. These models are capable
of producing text that resembles human writing, which is useful for converting large amounts of data into
readable summaries. When linked with robotic process automation tools such as UiPath Studio, LLMs can
adopt data-mining, summarization, and notification workflows.

This study aims to design and assess a technique that enables cross-platform notifications to be automated.
The system is designed to notify students, teachers, researchers, and anyone interested in events and their
details about the event through their email. This automated system is effective in improving the user
experience and ensuring the timely availability of information on events across various platforms.
 Literature Review
The rapidly evolving technology of robotic process automation (RPA) is designed to automate tasks,
jobs, and business processes in enterprises. Over time, RPA's scalability and efficiency in automating
repetitive tasks have garnered significant interest from both academic and industrial sectors.

Vinay [1], in his paper "Robotic Process Automation" published in the American Journal of Intelligent
Systems, examines the fundamentals of RPA. He emphasizes how the use of software robots, which
perform rule-based tasks with exceptional accuracy and efficiency, is transforming business process
automation. Vinay asserts that the increasing popularity of RPA is driven by its ability to enhance
operational efficiency and reduce human error in repetitive, high-volume tasks.

Potturu [2], in his study "UIPATH BOT FRAMEWORK: ACCELERATING RPA DEVELOPMENT AND
INNOVATION," highlights the role of UiPath in enhancing RPA implementation. He discusses how the
UiPath framework simplifies the adoption and innovation of RPA technology by accelerating the
development and deployment of RPA solutions. Potturu notes that the UiPath platform excels at
integrating with various enterprise systems and automating complex workflows.

Ketkar and Gawade [3], in their conference paper "Effectiveness of Robotic Process Automation for Data
Mining using UiPath," examine the accuracy and scalability of RPA in managing large datasets. They
demonstrate how UiPath bots can be used for effective data mining tasks, concluding that RPA, when
combined with data mining techniques, can significantly improve an organization's data analysis
capabilities, thereby enhancing decision-making processes.

Enríquez et al. [4] conducted an exhaustive mapping study on RPA, published in IEEE Access, providing
a comprehensive view of the current state of RPA research and applications. Their research indicates a
growing interest in validating RPA solutions in real-world scenarios, with a significant focus on the
development of RPA software platforms and theoretical models. Despite the widespread use of RPA in
many industries, the authors point out that more research is needed to fully understand its application
in complex, dynamic environments.

Khare et al. [5], in their paper "E-Mail Assistant – Automation of E-Mail Handling and Management
using Robotic Process Automation," presented at the International Conference on Decision Aid Sciences
and Applications, describe a novel application of RPA. They propose using RPA and artificial intelligence
(AI) to create an intelligent email assistant that automates email management tasks. Their analysis
suggests that combining AI with RPA, or cognitive automation, can enhance business process
performance and efficiency by handling unstructured data and making intelligent decisions.

Ribeiro et al. [6], in their paper "Robotic Process Automation and Artificial Intelligence in Industry 4.0,"
published in Procedia Computer Science, offer a thorough literature review on the convergence of RPA
and AI. They discuss how Industry 4.0 is evolving due to the integration of RPA and AI technologies,
enabling the creation of intelligent automation solutions that can adapt to complex industrial
environments. The authors emphasize that cognitive RPA can automate decision-making processes and
boost operational efficiency in various industries.
In summary, research on RPA demonstrates its significant impact on business process automation
through its scalability, accuracy, and efficiency. Scholars highlight the importance of integrating RPA
with AI to achieve cognitive automation, thereby enhancing the capability of software robots to handle
complex and dynamic tasks. Ongoing research and development in RPA technology are expected to
expand its utility and applications across diverse industrial settings.
This study leverages a combination of advanced tools and technologies to implement an automated system for
event notification, summarization, and dissemination. The primary tools and technologies used in this study
include:

1. UiPath Studio:
○ Web Scraping: UiPath Studio is employed for web scraping tasks. It utilizes the UI Explorer
feature to extract data from various event websites. The data extraction process involves
gathering information such as event titles, dates, locations, and descriptions.
○ Data Handling: The extracted data is stored in a structured format, specifically in a DataTable,
which facilitates further processing and summarization.
○ HTTP Requests: UiPath Studio’s HTTP request activities are used for sending automated
emails containing summarized event information to the users. This ensures timely and relevant
notifications.
2. Phi3 Language Model:
○ Data Summarization: The Phi3 Language Model, a lightweight Large Language Model
(LLM), is employed to convert large volumes of extracted data into readable and concise
summaries. This model is adept at producing human-like text, making the summarized data
more comprehensible and useful for users.
○ Natural Language Processing (NLP): The Phi3 Language Model utilizes NLP techniques to
analyze and transform raw data into meaningful summaries, enhancing the overall quality of
the information provided to the users.
3. HTTP Request for Email Notifications:
○ Automated Email Sending: The system uses HTTP request activities within UiPath Studio to
automate the process of sending emails. This integration ensures that the summarized event
information is dispatched to the concerned parties efficiently and without manual intervention.
4. DataTables:
○ Data Structuring: DataTables are used to store and manage the data extracted from event
websites. This structured format allows for easy manipulation, filtering, and processing of the
data, ensuring that the information is organized and readily available for summarization.
5. Gmail Activity in UiPath Studio:
○ Email Integration: The Gmail activity in UiPath Studio is utilized to send the automated
emails. This activity facilitates the integration of the email service with the automation
workflow, ensuring that users receive the event notifications in their inboxes promptly.

The combination of these tools and technologies provides a robust framework for automating the
identification, summarization, and notification of technical events. By leveraging UiPath Studio for web
scraping and email automation, the Phi3 Language Model for data summarization, and structured DataTables
for data management, the system ensures efficient and accurate delivery of relevant event information to the
users.
 Methodology
The methodology for this study involves a systematic approach to integrating Robotic Process Automation
(RPA) with Large Language Models (LLMs) to automate the identification, summarization, and notification
of technical events such as hackathons, coding competitions, and workshops. The process is divided into three
main stages: Data Extraction, Data Summarization, and Automated Email Notification.

A. System Architecture

The system architecture is designed to provide both cohesion and functionality across the entire framework. It
consists of three primary layers:

1. Data Extraction Layer:

○ Utilizes UiPath Studio to perform web scraping and collect event-related data from various
websites.
2. Data Processing and Summarization Layer:
○ Employs the Phi3 Language Model to transform raw data into readable summaries.
3. Notification Layer:
○ Uses UiPath Studio’s HTTP request activities to send email notifications containing
summarized event information to the users.

B. Workflow

The workflow of the system is well-planned, starting from data extraction to email notification upon
completion. The steps involved are as follows:

1. Data Extraction:
○ The system uses UiPath Studio to navigate to multiple event websites and retrieve event details
such as title, date, place, and description.
○ The extracted data is stored in a structured format, specifically a DataTable, for subsequent
processing.
2. Data Summarization:
○ The extracted data is converted into a string format necessary for the Phi3 Language Model.
○ The Phi3 Language Model processes the data to produce concise and readable summaries.
3. Automated Email Notification:
○ Using UiPath Studio’s HTTP request activities, the summarized data is sent as automated
emails to the specified recipients.
○ The system ensures that users receive timely and relevant notifications about upcoming events.

C. Detailed Steps

1. Data Extraction Process:

○ Provide the URL of the website from which data needs to be mined as input.
○ UiPath Studio performs web scraping to extract event information such as title, date, place, and
description.
 ○ Store the extracted data in a DataTable for further processing.
2. Data Processing and Summarization:
○ Convert the DataTable into a string format suitable for the Phi3 Language Model.
○ Construct the request body, including the extracted data and necessary parameters, and send it
to the Phi3 Model running on the local server through UiPath's SendRequest activity.
○ The Phi3 Model compiles the summarized data and produces a response, which is then cleaned
and formatted for compatibility with the email client.
3. Email Notification:
○ Use the cleaned and formatted summary to send automated emails to specified recipients using
the Gmail Activity in UiPath Studio.
○ Ensure the timely and relevant notification of users about upcoming events, enhancing their
experience by automating the notification process.

D. System Workflow Diagram

A diagram illustrating the system workflow includes the following steps:

1. Data extraction from event websites using UiPath Studio.

2. Data storage in a DataTable.
3. Data summarization using the Phi3 Language Model.
4. Sending summarized event information via automated emails using UiPath's HTTP request activities.

E. Efficiency and Accuracy

The methodology emphasizes efficiency and accuracy in data extraction, summarization, and notification. By
automating these processes, the system ensures that users receive timely and relevant information with
minimal manual intervention. The combination of RPA and LLMs provides a reliable solution for event
notification, improving user experience and ensuring that no significant opportunities are missed.

Overall, this methodology demonstrates how the integration of UiPath RPA and LLMs can effectively
automate the notification of technical events, ensuring that users stay informed and engaged with relevant
opportunities.
 Testing and Validation
The testing and validation of the proposed system involve several phases to ensure its accuracy,
efficiency, and reliability in identifying, summarizing, and notifying users about technical events. The
following steps outline the process of testing and validation:

A. Unit Testing

1. Data Extraction Testing:

○ Verify that the web scraping component in UiPath Studio accurately extracts event details
such as titles, dates, locations, and descriptions from various event websites.
○ Test the system with multiple URLs to ensure it can handle different website structures
and layouts.
2. Data Summarization Testing:
○ Test the Phi3 Language Model’s ability to summarize extracted data accurately and
coherently.
○ Compare the automated summaries with manually created summaries to evaluate the
quality and relevance of the information.
3. Email Notification Testing:
○ Ensure that the HTTP request activities in UiPath Studio correctly format and send
automated emails to specified recipients.
○ Verify that the emails contain the summarized event information and are delivered
promptly.

B. Integration Testing

1. Workflow Integration:
○ Test the entire workflow from data extraction to email notification to ensure seamless
integration between different components (UiPath Studio, Phi3 Language Model, and
email client).
○ Check for any data loss or errors during the transition between stages.
2. Error Handling:
○ Test the system’s ability to handle errors gracefully, such as missing data, website
changes, or network issues.
○ Ensure that appropriate error messages are logged and that the system can recover from
common failures.

C. Performance Testing

1. Scalability:
○ Evaluate the system’s performance when handling a large number of event websites and
data entries.
○ Measure the time taken for data extraction, summarization, and email notification to
ensure it meets performance requirements.
2. Efficiency:
○ Assess the system’s efficiency in terms of resource usage (CPU, memory) during the
execution of different tasks.
○ Optimize the system to minimize resource consumption while maintaining high
performance.
 D. User Acceptance Testing (UAT)

1. User Feedback:
○ Conduct testing sessions with a group of end-users (students, teachers, researchers) to
gather feedback on the system’s usability and functionality.
○ Ensure that the system meets user expectations for ease of use, accuracy of information,
and timely notifications.
2. Usability Testing:
○ Test the user interface and interaction flow to ensure that users can easily understand
and navigate the system.
○ Make necessary adjustments based on user feedback to enhance the overall user
experience.

E. Validation

1. Accuracy Validation:
○ Validate the accuracy of the data extracted and summarized by comparing it with the
original event information on the websites.
○ Use metrics such as precision, recall, and F1-score to quantify the accuracy of the
summarization process.
2. Effectiveness Validation:
○ Measure the effectiveness of the automated email notifications by tracking user
engagement (e.g., email open rates, click-through rates).
○ Gather feedback from users on the relevance and usefulness of the event information
provided.

F. Results Analysis

1. Comparison with Manual Process:

○ Compare the automated system’s performance with the manual process of identifying,
summarizing, and notifying events.
○ Highlight the time savings, accuracy improvements, and reduced manual effort achieved
through automation.
2. Continuous Improvement:
○ Analyze the test results and identify areas for improvement.
○ Implement iterative updates to the system based on testing outcomes and user feedback
to enhance its performance and reliability.
 Results and Discussion

The results of this study highlight the effectiveness of integrating UiPath Robotic Process Automation (RPA)
with the Phi3 Language Model to automate the identification, summarization, and notification of technical
events such as hackathons, coding competitions, and workshops. The following sections provide a detailed
analysis of the outcomes and their implications.

A. Data Extraction Results

1. Accuracy:
○ The web scraping component of UiPath Studio successfully extracted event details from
multiple websites with high accuracy.
○ The extracted data included event titles, dates, locations, and descriptions, stored in a
structured DataTable format.
2. Efficiency:
○ The automated data extraction process significantly reduced the time and effort required
compared to manual extraction methods.
○ UiPath Studio's ability to handle various website structures and layouts ensured comprehensive
data collection across different sources.

B. Data Summarization Results

1. Quality of Summaries:
○ The Phi3 Language Model generated concise and readable summaries from the extracted event
data.
○ A comparison between automated and manually created summaries showed a high degree of
relevance and coherence, demonstrating the model’s effectiveness in summarizing large
volumes of data.
2. User Feedback:
○ Users reported that the summaries provided essential information without overwhelming them
with unnecessary details.
○ The automated summaries were found to be as informative as manually written ones,
indicating the model’s capability in producing human-like text.

C. Automated Email Notification Results

1. Timeliness:
○ The system successfully sent automated email notifications containing summarized event
information to users in a timely manner.
○ The use of UiPath Studio’s HTTP request activities ensured that notifications were dispatched
promptly after data processing.
2. User Engagement:
○ Metrics such as email open rates and click-through rates indicated a high level of user
 engagement with the notifications.
○ Users appreciated the timely and relevant information provided, which helped them stay
informed about upcoming events without actively searching for them.

D. Performance Analysis

1. Scalability:
○ The system demonstrated scalability by efficiently handling a large number of event websites
and data entries.
○ Performance tests showed that the system maintained high accuracy and efficiency even under
increased workloads.
2. Resource Utilization:
○ The system’s resource usage (CPU, memory) was optimized to ensure minimal impact on
overall performance.
○ The integration of RPA and LLMs proved to be resource-efficient, making the system suitable
for continuous operation.

E. Comparative Analysis

1. Manual vs. Automated Process:

○ The automated system outperformed the manual process in terms of speed, accuracy, and
consistency.
○ Users saved considerable time and effort by relying on the automated notifications instead of
manually searching for event information.
2. Error Reduction:
○ The automated system minimized human errors in data extraction, summarization, and
notification, ensuring higher reliability and accuracy.
○ The consistency of the automated process reduced the chances of missing important events.

F. Discussion

1. Implications for Users:

○ The integration of RPA and LLMs provided a reliable and efficient solution for students,
teachers, researchers, and other users interested in technical events.
○ The system’s ability to deliver timely and relevant notifications improved users’ awareness and
participation in events, enhancing their professional and academic opportunities.
2. Technical Considerations:
○ The study demonstrated the feasibility of combining RPA with LLMs to automate complex
workflows involving data extraction, summarization, and notification.
○ The choice of UiPath Studio for RPA and the Phi3 Language Model for NLP was validated
through successful implementation and positive user feedback.
3. Future Enhancements:
○ The system can be further improved by incorporating additional features such as user-specific
customization, event categorization, and integration with more data sources.
○ Continuous updates to the Phi3 Language Model and UiPath Studio can enhance the system’s
accuracy and efficiency over time.
 Conclusion
By integrating UiPath Robotic Process Automation (RPA) with large language models including
the Phi3 Language Model, key technical activities like a hackathon or a workshop can be
efficiently identified, summarized, and even notified to end-users. This integration minimizes
the work done by administrators in data procurement, analysis, and alerting, while optimizing
the flow of information to users without irritating them with floods of notifications. Data input
is faster, the summary is accurate and automated notification is highly reliable thus makes the
system to ease operational activities and minimize on cases where important opportunities may
not be noticed. Therefore, there are clear implications for scaling up replication work across
most disciplines through smart automation system because the subject under study may require
further for developing effective techniques of enhancing users’ productivity and satisfaction.
 References
Vinay, Kommera. ”Robotic Process Automation.” American Journal of
Intelligent Systems, 9(2):49-53, 2019. doi: 10.5923/J.AJIS.20190902.01.

https://arxiv.org/abs/2205.05882

Y. Ketkar and S. Gawade. ”Effectiveness of Robotic Process

Automation for data mining using UiPath.” 2021 International
Conference on Artificial Intelligence and Smart Systems (ICAIS),
Coimbatore, India, 2021, pp. 864-867. doi:
10.1109/ICAIS50930.2021.9396024

https://arxiv.org/abs/2205.05882