Digital humanities integrate computational tools into traditional humanities disciplines to explore new research methodologies. The incorporation of advanced digital analysis and LLMs promises substantial enhancements in text analysis, data visualization, and cultural data analytics. This paper aims to systematically evaluate these impacts, offering insights into both the advancements and complications posed by these technologies. Digital humanities encompass a multidisciplinary field where digital technology intersects with arts, humanities, and social sciences research. The advent of advanced AI technologies, especially LLMs like OpenAI's GPT series, has introduced new tools for textual analysis, data interpretation, and even the generation of human-like text, offering both unprecedented opportunities and significant challenges. This paper explores these dynamics, offering insights into the integration of LLMs in DH and discussing the broader implications for researchers.
Landscape represents the interconnected technologies, methodologies, and considerations that shape the field of DH [1]. Each of these areas contributes to how humanities scholars engage with digital tools and data to conduct research, preserve cultural heritage, and disseminate knowledge. We can distinguish the following components of a landscape relating to the DH:
• Core Technologies: Artificial Intelligence and Machine Learning [2,3], Neural networks and trasformers [4,5], Natural Language Processing [6], Big Data Analytics [7,8], Cloud Computing [9,10], Blockchain. • Research Tools: Text Analysis Software, Data Visualization Tools, Geographic Information Systems(GIS), Digital Asset Management Systems, Collaborative Platforms.
ICYRIME 2024: 9th International Conference of Yearly Reports on Informatics, Mathematics, and Engineering. Catania, July 29-August 1, 2024 a.cigliano@unimarconi.it (A. Cigliano); f.fallucchi@unimarconi.it (F. Fallucchi); m.gerardi@unimarconi.it (M. Gerardi) in Cultural Heritage, Quantum Computing Applications, Non-Terrestrial Networks (NTNs) for monitoring applications [11], Advanced Natural Language Generation (LLMs).
Digital analysis involves the use of software and algorithms to analyze cultural and historical data. It includes techniques like data mining, visualization, and statistical analysis, which allow scholars to uncover patterns and trends that are not apparent through traditional methods.
LLMs like OpenAI's GPT-3 represent a breakthrough in natural language processing technology from November 2022. These models, trained on extensive datasets, are capable of generating coherent text, completing linguistic tasks, and providing new tools for textual interpretation and generation in the humanities. Generative AI refers to algorithms capable of creating content, from text to images, by learning
This study synthesizes data from various case studies, peerreviewed articles, and firsthand experiments with LLMs in digital humanities projects. Through qualitative and quantitative analysis [13], we assess how these technologies are currently applied and theorize their future trajectories within the field.
The applications of digital analysis tools and Large Language Models (LLMs) in digital humanities are vast and diverse. These technologies have revolutionized how researchers, educators, and practitioners approach the humanities, offering new insights and methodologies that were previously unattainable. Below is an exploration of several key applications and how they impact the field of digital humanities.
This is particularly useful in experimental literature, automated storytelling, and the creation of dialogue or narrative for digital artifacts [14]. These models are adept at interpreting and providing context to vast archives of text, helping scholars identify themes, trends, and patterns across large datasets that would be unmanageable for manual review. In this context we can distinguish the following subapplications:
• Semantic Analysis: LLMs can analyze large bodies of text to understand contextual meanings, identify themes, and detect nuances in language that are often missed by traditional analysis methods. This capability is crucial for Literary analysis, historical documentation, and cultural studies. • Sentiment Analysis: Digital tools can assess the sentiment of historical texts, literary works, and even large collections of social media posts to gauge public sentiments over time or reactions to specific events or figures in history. • Stylistic Analysis: LLMs are used to study stylistic elements of writing across different time periods or authors. This application helps in authorship attribution and understanding the evolution of language and writing styles in literature.
We can distinguish two macro components: Pattern Recognition, in this case Digital analysis tools can sift through vast amounts of data to identify patterns and trends [15]. In historical studies, for example, this can reveal migration patterns, economic changes, or social dynamics of a particular era. Network Analysis, where in digital humanities, network analysis tools can map relationships among various entities within literature, such as characters in novels, historical figures, or even concepts and ideas that are prevalent within a specific cultural or intellectual community.
Machine Translation in LLMs models, provide tools for translating rare or endangered languages, making ancient manuscripts or texts accessible to a global audience. This application is particularly significant in preserving linguistic heritage and making non-English academic resources available to a broader audience. Another case is Automatic Transcription, where LLMs and digital tools automate the transcription of audio recordings into text, such as oral histories and lectures, which can then be analyzed or archived digitally for future research. In general LLMs and GenAI aids in transcribing handwritten or archaic documents into digital formats, which are then more accessible for analysis and interpretation(Historical Text Transcription). AI models facilitate the translation of texts across languages, making non-native and ancient literature accessible to a global audience without the immediate need for human translators(Translation).
We can distinguish Document Digitization and Categorization, in this case Digital tools automate the scanning and categorization of physical documents into digital archives. LLMs can enhance this process by automatically tagging and organizing the documents based on their content. Preservation of Digital Artifacts: Digital humanities also involve the preservation of digital art and online culture, where digital analysis tools help in archiving web-based art and social media content.
Understanding the deeper meanings, connotations, and contexts of words in historical and contemporary texts(Semantic Analysis). Analyzing the emotions or sentiments expressed in texts, which can be particularly useful in studies of literature, social media, and historical documents to gauge public opinion and cultural trends over time(Sentiment Analysis).
AI can generate dynamic visual representations of textual data, making it easier for researchers to explore complex datasets and uncover relationships (Interactive Data Displays). Linking textual data with geographical information systems (GIS) to map the occurrence of certain themes or discussions across different locations(Geospatial Mapping).
In digital humanities, the classification of AI tools also involves identifying potential biases in AI-generated content and developing methodologies to mitigate these biases (Bias Mitigation) [16]. Using AI to verify the authenticity of digital artifacts and trace their historical and cultural origins(Authenticity and Provenance Verification).
LLMs identify and analyze patterns in art, music, literature, and other cultural artifacts over large temporal and spatial scales (Cultural Pattern Recognition). This includes using AI to critique and analyze its own impact on digital humanities, assessing how technology shapes cultural and societal understanding(Critical AI Studies).
Digital Archiving: AI facilitates the organization, categorization, and searchability of digital archives, enhancing the accessibility of vast digital collections. Curatorial Assistance: AI tools assist in curating digital exhibitions and virtual museum tours by selecting themes and artifacts based on scholarly input and visitor interactions. These categories reflect the multifaceted roles that generative AI and LLMs play in the digital humanities [17,18]. They not only enhance traditional research methodologies but also introduce new modes of scholarly inquiry and interaction with digital texts and datasets. As these technologies evolve, their classification may expand to include more specialized functions and applications tailored to the nuanced needs of humanities research.
Interactive Exhibits: Museums and educational institutions utilize digital tools to create virtual tours and interactive exhibits that allow users to explore artifacts, artworks, and historical documents from their devices, enhancing accessibility and engagement. Virtual Reality (VR) and Augmented Reality (AR): These technologies are employed to create immersive educational experiences, such as virtual reality environments that simulate historical sites or events [19].
Adaptive Learning Systems: Digital analysis tools can help develop educational platforms that adapt to the learning pace and style of individual students, particularly in humanities education, by analyzing student interactions and providing customized feedback. Online Courses and Workshops: LLMs can assist in the creation of educational content, generating reading materials, quizzes, and interactive discussions that facilitate online learning in the humanities.
Social Media Analysis: Digital tools analyze social media data to study contemporary cultural trends and public opinions, providing humanities researchers with real-time data on human behavior and societal changes. Crowdsourcing and Collaborative Research: Platforms designed with the help of digital tools enable collaborative projects where scholars and the public can contribute to and engage with humanities research globally.
Digital humanities involve applying digital tools to humanities subjects, enabling new research methodologies like text mining, machine learning, and data visualization. These tools facilitate the exploration of cultural patterns, historical data, and literary analysis at scales previously unmanageable. DH is an interdisciplinary field that merges the techniques and methodologies of the digital world with the study of the humanities. This field includes, but is not limited to, disciplines like literature, history, art, music, philosophy, cultural studies, and linguistics. The scope of DH is vast, integrating computational methods into the traditional humanities to enhance research, teaching, and the dissemination of knowledge.
• Textual Analysis: Digital tools allow for the analysis of large volumes of text, enabling scholars to uncover patterns, trends, and insights at a scale not feasible manually; • Historical GIS (Geographic Information Systems): Integrates geographical data into humanities research, allowing for the spatial analysis of historical events and cultural developments. • Cultural Analytics: Uses data analysis to study patterns in visual culture, music, and other art forms, identifying trends and influences over time.
• Promotes collaboration across disciplines, combining methods from social sciences, computer sciences, and humanities to tackle complex research questions. • Encourages partnerships between academicians, technologists, and sometimes the public, fostering a richer, more diverse research environment, including emulation platforms [20].
• Digital exhibitions and virtual reality experiences make humanities more accessible to the public. • Open-source projects and tools democratize access to knowledge, allowing broader participation in humanities scholarship.
Digital archiving ensures the preservation of important historical and cultural materials, making them accessible worldwide and safeguarding them against physical degradation.
Digital platforms facilitate the sharing of research outputs and scholarly discourse more dynamically and widely than traditional publishing methods.
The tools used in digital humanities are diverse, ranging from data analysis software to digital storytelling platforms, each serving different aspects of humanities research: 1. Text Analysis Tools: -Natural Language Processing (NLP) Software: Tools like NLTK, spaCy, and GPT-3 help analyze text for syntax, sentiment, and semantic content.
-Concordance Software: Allows scholars to locate every occurrence of a word or phrase, facilitating analysis of texts.
2. Data Visualization Tools: -Tableau, Gephi: Useful for creating visual representations of data, helping to illustrate complex relationships and patterns. -Timeline Tools: Tools like TimelineJS help researchers create interactive timelines that highlight historical events and trends.
3. Mapping Tools: -GIS Software: ArcGIS and QGIS are used to analyze and display spatial data, providing insights into geographical aspects of historical and cultural studies. -StoryMap JS: Allows for the creation of maps that combine narrative text with images and multimedia content.
4. Digital Archiving Systems: -Omeka: Provides a web platform for publishing collections and exhibitions, widely used by libraries, museums, and archives. -Digital Asset Management Systems (DAMS): Helps institutions manage their digital content effectively.
5. Programming and Development Environments: -Python and R: Popular programming languages for handling data-intensive projects in DH. -Jupyter Notebooks: Offers an interactive environment for coding, visualizing, and publishing data analysis projects.
6. Collaborative Tools: -Wiki Software: Facilitates collaborative writing and knowledge sharing. -Slack, GitHub: Used for project management and collaboration in larger digital projects.
7. Content Management Systems (CMS): -WordPress, Drupal: Often used to create and manage digital content, facilitating dynamic web presence for DH projects.
The integration of these tools into humanities research has not only transformed traditional methodologies but also expanded the boundaries of what can be explored and understood in the humanities. As digital technologies continue to evolve, the scope and tools of digital humanities will likely expand further, continuing to redefine the engagement with humanistic studies in the digital age.
This study employs a qualitative research approach, synthesizing information from various case studies, scholarly articles, and empirical research findings to evaluate the impact of LLMs in digital humanities. The research includes a meta-analysis of published digital humanities projects utilizing generative AI and a review of literature discussing theoretical and methodological implications.
LLMs are particularly potent in their ability to process and analyze large volumes of text, providing insights into linguistic patterns, stylistic features, and historical language evolution. Examples include the use of AI for analyzing vast archives of unstructured text in libraries and museums. Text analysis and interpretation form a core component of the intersection between Large Language Models (LLMs) and digital humanities. LLMs, due to their deep learning foundations and substantial training datasets, offer sophisticated tools for automating and enhancing the examination of textual data. This synergy is particularly influential in the field of digital humanities, where traditional hermeneutic methods meet modern computational analytics. Text analysis involves using algorithms to understand, interpret, and
derive meaningful information from textual data. In the context of digital humanities, LLMs apply several techniques:
• Tokenization and Parsing: Breaking down text into manageable pieces or tokens to analyze structural and grammatical relationships. • Named Entity Recognition (NER): Automatically identifying and classifying key elements in text into predefined categories such as names of people, organizations, locations, etc. • Sentiment Analysis: Determining the affective state of the text, whether positive, negative, or neutral, which can be particularly useful in analyzing literary works or historical documents to gauge the sentiments expressed.
Generative AI assists in the transcription and translation of historical manuscripts, making them more accessible to researchers worldwide. This not only democratizes access but also enriches global understanding of historical contexts.
In creative domains, LLMs have been used to generate literary texts and assist in experimental poetry and narrative projects, often blurring the lines between human and machine creativity.
Regarding the challenges and limitations we can identify the following strands: Technological Dependence, the reliance on sophisticated computational tools requires ongoing technical support and significant resources, which can be a barrier for institutions with limited funding. Data Privacy and Ethical Concerns, as with all digital data, there are concerns regarding privacy, particularly when dealing with sensitive or personal historical records. Ethical considerations also arise in how data is interpreted and presented [21]. The Impact on research and scholarship is relevant, in particular scale and Scope: Digital tools allow researchers to handle datasets of unprecedented size and complexity, often referred to as "big data" in humanities [22]. This capacity has led to the emergence of "distant reading" techniques, where scholars analyze patterns across thousands of texts, rather than studying a single piece in depth ("close reading").
Concerns about data privacy and the ethical use of digital archives are paramount, particularly when dealing with sensitive or personal historical data.
The reliance on proprietary AI technologies, often developed by major corporations, poses risks related to accessibility, transparency, and continuity of research. [23]. The "black box" nature of many AI systems makes it challenging to understand how they arrive at certain conclusions or interpretations [24,25]. This lack of transparency can be problematic in humanities research, where understanding the reasoning behind interpretations is crucial. Using AI to analyze historical or cultural data raises questions about consent, especially when dealing with sensitive or personal information from the past [26]. There's a need to balance research benefits with respect for privacy and cultural sensitivities. As AI becomes more capable of generating human-like text, there are concerns about maintaining the authenticity of historical documents and distinguishing between AI-generated and human-created content. Over-reliance on AI tools could potentially lead to a reduction in critical thinking skills or traditional research methods in humanities [27]. There's a need to find a balance between leveraging AI capabilities and maintaining core humanities research skills. Researchers must consider the ethical implications of using AI-generated insights, especially when these insights could impact cultural understanding or policy decisions. The adoption of AI tools in digital humanities could create or exacerbate existing inequalities between well-funded institutions and those with fewer resources [28]. This raises questions about equitable access to advanced research tools and methodologies. AI systems may not always understand or respect cultural nuances, leading to misinterpretations or insensitive handling of cultural data. Determining responsibility 1-7
for errors or biased outcomes in AI-assisted research can be challenging [29]. There's a need for clear guidelines on accountability in collaborative human-AI research projects.
To address these concerns, researchers and institutions can:
In conclusion, digital analysis within digital humanities is not just about incorporating new tools but fundamentally transforming the ways that humanistic research can be conducted, presented, and taught. This evolution continues to offer both exciting opportunities and significant challenges that will shape the future of humanities scholarship. Generative AI and LLMs offer transformative potential for digital humanities, enabling sophisticated textual analysis and broadening the scope of what is technologically possible in humanities research. However, their integration must be handled with care, considering ethical, methodological, and practical challenges. Future research should focus on developing equitable and transparent use of these technologies to enhance the digital humanities field.