Journal of Artificial Intelligence and Machine Learning (JAIML)

Volume 3, Issue 1, January-June 2024, pp. 1-7, Article ID: JAIML_03_01_001

Available online at https://iaeme.com/Home/issue/JAIML?Volume=3&Issue=1
Journal ID: 5979-A561

© IAEME Publication

TEXT SUMMARIZATION USING GENERATIVE

AI: A CASE STUDY IN BANKING INDUSTRY
Karthika Gopalakrishnan
Data Scientist, United States

ABSTRACT
This paper presents a novel approach to text summarization utilizing Generative AI,
specifically leveraging OpenAI and Lang Chain technologies. We introduce the
architecture of the summarization application, demonstrate its implementation with
OpenAI's Language Model (LLM), and elucidate the process of text segmentation and
summarization. Furthermore, the study discusses the potential applications of text
summarization in the banking industry, focusing on simplifying manual tasks such as
summarizing multi-page loan applications or account opening forms. Finally, the paper
addresses the importance of caution in deploying Generative AI solutions in the banking
sector, emphasizing the need for robust security measures and ethical considerations.
Keywords: Text Summarization, Generative AI, OpenAI, Lang Chain, Banking
Industry, Language Model.
Cite this Article: Karthika Gopalakrishnan, Text Summarization Using Generative AI:
A Case Study in Banking Industry, Journal of Artificial Intelligence and Machine
Learning (JAIML), 3(1), 2024, pp. 1-7
https://iaeme.com/Home/issue/JAIML?Volume=3&Issue=1

1. INTRODUCTION
In recent years, the exponential growth of textual data has been unparalleled, fueled by the
proliferation of digital platforms, social media, and online communication channels. This
deluge of text-based information presents significant challenges in information processing and
comprehension. Individuals and organizations are inundated with vast amounts of text, ranging
from articles, reports, emails, to social media posts and messages. Amidst this abundance, the
need for efficient methods to extract actionable insights and distill key information becomes
paramount.
Text summarization emerges as a critical natural language processing (NLP) task aimed at
addressing this challenge. At its core, text summarization seeks to condense lengthy documents
or passages into concise summaries while preserving the essential meaning and context. By
extracting salient information and discarding redundant or extraneous details, summarization
facilitates quicker understanding, decision-making, and information retrieval.
Traditional methods of text summarization often rely on heuristics, statistical models, or
rule-based approaches. While these methods can be effective to some extent, they frequently
struggle to capture the semantic nuances and intricacies of human language.

https://iaeme.com/Home/journal/JAIML 1 [email protected]
 Text Summarization Using Generative AI: A Case Study in Banking Industry

As a result, the generated summaries may lack coherence, miss important details, or fail to
convey the intended meaning accurately.
Generative AI, powered by deep learning models, offers a compelling solution to enhance
text summarization capabilities. These models leverage neural network architectures, such as
recurrent neural networks (RNNs), convolutional neural networks (CNNs), or transformer
models, to learn intricate patterns and structures within textual data. By training on vast corpora
of text, generative AI models can develop a nuanced understanding of language semantics,
syntax, and context.
One of the most notable advancements in generative AI is the development of transformer-
based models, exemplified by OpenAI's Language Model (LLM). These models excel in natural
language understanding and generation, exhibiting remarkable proficiency in various NLP
tasks, including text summarization. Through the process of unsupervised learning, transformer
models can effectively capture long-range dependencies, semantic relationships, and contextual
nuances present in textual data.
By harnessing the power of generative AI, text summarization systems can produce
summaries that are more coherent, contextually relevant, and informative. These systems have
the potential to revolutionize information processing workflows across industries, enabling
faster decision-making, knowledge extraction, and content consumption.
In summary, the exponential growth of textual data necessitates innovative solutions to
extract meaningful insights and distill key information effectively. Text summarization, as a
crucial NLP task, addresses this challenge by condensing lengthy documents into concise
summaries. Generative AI, powered by deep learning models like transformer architectures,
holds immense promise in enhancing text summarization capabilities by capturing semantic
nuances and generating coherent summaries.

2. REVOLUTIONIZING TEXT SUMMARIZATION: THE POWER OF

GENERATIVE AI
Text summarization has undergone a dramatic transformation with the emergence of generative
AI. This innovative approach surpasses traditional methods that merely reassemble existing
sentences. Generative AI models, with their remarkable ability to create entirely new content,
offer a sophisticated way to extract key information from textual data.
At the core of generative AI lies its ability to mimic human language comprehension and
manipulation. These models, frequently built on transformer architectures like OpenAI's GPT
series, are trained on massive datasets encompassing a vast array of text sources. This training
process empowers them to understand the intricate relationships between words, phrases, and
sentences, along with the subtle contextual cues that breathe meaning into language.
When applied to text summarization, generative AI meticulously analyzes the input text
before crafting a concise version that captures the essence of the information. Unlike extractive
summarization, which relies on piecing together existing sentences, generative AI models can
generate entirely new sentences that convey the core message in a clear and succinct way.
The true strength of generative AI in text summarization lies in its ability to produce
summaries that exhibit a human-like grasp of the source material. By drawing upon learned
linguistic patterns and semantic structures, these models can create summaries that not only
capture the critical points but also preserve the context and tone of the original content. This
results in summaries that are not just informative but also engaging and stylistically consistent
with human-written text.

https://iaeme.com/Home/journal/JAIML 2 [email protected]
 Karthika Gopalakrishnan

Furthermore, generative AI paves the way for summarization beyond simple extraction. It
allows for the synthesis of information from various sources or the generation of abstractive
summaries that delve deeper than surface-level details. This capability is particularly valuable
for complex or ambiguous texts, as generative AI models can infer implicit meaning and
produce summaries that capture the underlying essence of the content.
Generative AI for text summarization marks a significant leap forward in our pursuit of
advanced natural language understanding and generation. Leveraging the power of deep
learning and transformer architectures, these models offer a versatile and powerful tool for
transforming complex textual information into concise and meaningful summaries, impacting
a multitude of applications across diverse industries.

3. OPENAI AND LANG CHAIN

3.1. OpenAI
OpenAI is at the forefront of artificial intelligence research, developing cutting-edge
technologies to advance machine learning capabilities. One of its flagship offerings is the
Language Model (LLM), a powerful generative model trained on vast amounts of textual data.
LLM excels in natural language understanding and generation, making it an ideal candidate for
various NLP tasks, including text summarization.

3.2. Lang Chain

Lang Chain is a framework that integrates OpenAI's LLM into applications, facilitating
seamless integration and utilization of generative AI capabilities. It provides developers with a
robust toolkit for building sophisticated NLP applications, including text summarization
systems. Lang Chain abstracts the complexities of model deployment and management,
enabling efficient development and deployment of AI-powered solutions.

4. TEXT SUMMARIZATION IMPLEMENTATION

The study creates a Text Summarization application using OpenAI and Langchain. The
application is hosted in Streamlit to provide interactive user interface.

4.1. LLM Model Initialization

Open AI’s LLM models are accessed through its libraries like openai. The specific LLM choice
depends on factors like document complexity and desired summary length.

4.2. Text Splitting

For long documents, the text is split into smaller chunks using LangChain's splitting functions.
This improves efficiency and avoids overwhelming the LLM.

4.3. Text Summarization

Each text chunk is sent to the LLM for summarization. The LLM generates a concise summary
capturing the key points. LangChain provides “load_summarize_chain” to summarize multiple
documents. The study uses “map_reduce” summarization chain.

Map-Reduce for Text Summarization:

Map: The input text (if long) is divided into smaller chunks. Each chunk is then "mapped" to
the LLM for individual summarization.

https://iaeme.com/Home/journal/JAIML 3 [email protected]
 Text Summarization Using Generative AI: A Case Study in Banking Industry

Reduce: The individual summaries from each chunk are combined into a single, concise
summary of the entire document.
The benefit of using a map-reduce approach lies in handling long documents efficiently. By
processing smaller chunks, you avoid overwhelming the LLM and potentially improve the
overall quality of the summary.
Figure 1 shows the workflow of the Text Summarization application and Figure 2 shows
the Text Summarization application hosted in streamlit. Figure 3 shows the sample input
(Business information from 10K report – IBM Corporation) and summarized output produced
by the Text Summarization application. IBM

Figure 1: Text Summarization Workflow

Figure 2: Text Summarization Application

https://iaeme.com/Home/journal/JAIML 4 [email protected]
 Karthika Gopalakrishnan

Figure 3: Sample Output - Text Summarization

5. TEXT SUMMARIZATION WITH GENERATIVE AI IN BANKING:

APPLICATIONS AND BENEFITS
In the banking industry, text summarization using generative AI presents numerous
opportunities to streamline processes and enhance efficiency. Here are some specific
applications:

5.1. Loan Application Summarization

Problem: Loan applications often contain extensive information ranging from financial details
to employment history, making the review process time-consuming for loan officers.
Solution: Generative AI can be employed to generate concise summaries of loan applications,
focusing on key financial metrics such as income, credit score, debt-to-income ratio, and loan
amount. These summaries enable loan officers to quickly assess applicant profiles, accelerating
decision-making processes and improving operational efficiency.

5.2. Account Opening Form Summarization

Problem: Account opening forms typically include repetitive information like customer details
and investment goals, making manual extraction tedious and error prone.
Solution: By leveraging Generative AI, banks can automatically summarize account opening
forms to extract essential details such as customer information, contact details, and investment
preferences. This streamlines the account opening process, reducing administrative burdens for
both customers and bank staff.

5.3. Customer Service Interaction Summarization

Problem: Analyzing large volumes of customer service interactions to identify trends and
recurring issues is time-consuming and resource intensive.
Solution: Generative AI can be utilized to summarize customer service interactions, identifying
common concerns and feedback themes. These summaries empower banks to gain insights into
customer preferences and pain points, enabling them to enhance their service offerings and
improve customer satisfaction levels.

5.4. Regulatory Compliance Review Summarization

Problem: Reviewing extensive regulatory documents and reports to extract compliance
requirements and identify potential risks is challenging and prone to human error.

https://iaeme.com/Home/journal/JAIML 5 [email protected]
 Text Summarization Using Generative AI: A Case Study in Banking Industry

Solution: Generative AI-powered text summarization can automate the summarization of

regulatory documents, highlighting essential compliance points and potential risk areas. This
facilitates faster and more accurate compliance reviews, ensuring regulatory adherence and
mitigating compliance-related risks.

5.5. Internal Document Summarization

Problem: Banks generate a plethora of internal documents such as meeting minutes, project
reports, and strategic plans, making information retrieval and knowledge sharing cumbersome.
Solution: By employing Generative AI for text summarization, banks can summarize internal
documents, facilitating knowledge sharing and improving information retrieval processes.
These summaries enable employees to quickly access relevant information, fostering
collaboration and informed decision-making across departments.
Generative AI-powered text summarization presents a myriad of opportunities for the
banking industry to enhance operational efficiency, improve customer service, ensure
regulatory compliance, and streamline internal processes. By leveraging advanced AI
technologies, banks can unlock valuable insights from vast volumes of textual data, driving
innovation and differentiation in an increasingly competitive landscape.

6. RESPONSIBLE USE OF AI
While text summarization offers numerous benefits in the banking sector, there are inherent
risks that must be addressed:

6.1. Data Privacy

Banking documents contain sensitive information, necessitating robust data protection
measures to safeguard customer privacy.

6.2. Model Bias

Generative AI models may exhibit biases present in the training data, leading to skewed or
inaccurate summaries. Regular monitoring and mitigation of biases are essential to ensure fair
and unbiased summarizations.

6.3. Security Concerns

Deploying AI-powered systems introduces security vulnerabilities, such as adversarial attacks
and model exploits. Banks must implement stringent security protocols to mitigate these risks
and protect against potential breaches.

7. CONCLUSION
Text summarization using Generative AI presents a transformative solution for simplifying
manual tasks and enhancing operational efficiency in the banking industry. By leveraging
OpenAI's Language Model and Lang Chain framework, banks can develop robust
summarization systems capable of handling complex documents with ease. However, cautious
implementation and adherence to ethical guidelines are imperative to mitigate risks and ensure
the responsible deployment of AI technologies in banking operations. Further research can
explore integrating the text summarization application with existing banking workflows.
Additionally, exploring explainable AI techniques can provide insights into the LLM's
reasoning behind generated summaries, fostering trust and transparency.

https://iaeme.com/Home/journal/JAIML 6 [email protected]
 Karthika Gopalakrishnan

REFERENCES
[1] N. K. Shukla et al., "Generative AI Approach to Distributed Summarization of Financial
Narratives," 2023 IEEE International Conference on Big Data (BigData), Sorrento, Italy, 2023,
pp. 2872-2876, doi: 10.1109/BigData59044.2023.10386313.

[2] P. Raundale and H. Shekhar, "Analytical study of Text Summarization Techniques," 2021 Asian
Conference on Innovation in Technology (ASIANCON), PUNE, India, 2021, pp. 1-4, doi:
10.1109/ASIANCON51346.2021.9544804.

[3] Roy, Kunal & Mukherjee, Subhash & Dawn, Sujata. (2023). Automated Article Summarization
using Artificial Intelligence Using React JS and Generative AI.

[4] OpenAI. (n.d.). OpenAI Language Model. https://openai.com/language-models

[5] Lang Chain. (n.d.). Lang Chain Documentation. https://langchain.com/documentation

Citation: Karthika Gopalakrishnan, Text Summarization Using Generative AI: A Case Study in Banking Industry,
Journal of Artificial Intelligence and Machine Learning (JAIML), 3(1), 2024, pp. 1-7

Abstract Link:
https://iaeme.com/Home/article_id/JAIML_03_01_001

Article Link:
https://iaeme.com/MasterAdmin/Journal_uploads/JAIML/VOLUME_3_ISSUE_1/JAIML_03_01_001.pdf

Copyright: © 2024 Authors. This is an open-access article distributed under the terms of the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original author and source are credited.

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).

✉ [email protected]

https://iaeme.com/Home/journal/JAIML 7 [email protected]