This paper introduces Perspectives, an interactive extension of the Discourse Analysis Tool Suite designed to empower Digital Humanities (DH) scholars to explore and organize large, unstructured document collections. Perspectives implements a flexible, aspect-focused document clustering pipeline with human-in-the-loop refinement capabilities. We showcase how this process can be initially steered by defining analytical lenses through document rewriting prompts and instruction-based embeddings, and further aligned with user intent through tools for refining clusters and mechanisms for fine-tuning the embedding model. The demonstration highlights a typical workflow, illustrating how DH researchers can leverage Perspectives's interactive document map to uncover topics, sentiments, or other relevant categories, thereby gaining insights and preparing their data for subsequent in-depth analysis.
Figures & tables
Figure 1: Perspectives’ document map. Center: interactive scatter plot of documents colored by cluster. Hovering over a document previews its content. Left: settings & refinement operations, including the model fine-tuning operation. Right: statistics & information about selected documents. Top: toolbar with search & filtering. The UI is inspired by popular clustering interfaces such as Nomic Atlas, leveraging a familiar design to accelerate adoption.
Figure 2: The proposed interactive clustering pipeline. The initial clustering process is guided by providing rewriting and embedding instructions (green) to focus the document representations on user-defined aspects. The established core pipeline (orange) identifies clusters and builds various textual representations. Users can post-process the clustering (blue) through refinement operations (grey), triggering some steps of the pipeline.
emotion
genre
product
stars
topic
frames
blurbs
stance
bias
text
45.10
27.20
36.72
47.54
71.17
59.78
73.84
51.19
47.20
+inst
50.60
27.50
55.76
58.86
71.24
65.00
73.25
48.21
48.51
keyphrases
49.15
34.85
54.64
46.34
65.70
60.34
76.73
61.63
50.76
+inst
49.51
34.90
60.64
49.68
65.40
58.31
76.66
63.14
50.69
summary
47.69
32.59
48.16
52.14
60.62
64.34
76.46
62.47
48.41
+inst
48.04
32.64
61.18
54.02
61.91
65.91
74.80
68.74
52.53
Table 1: Evaluation of document rewriting, instruction-based embeddings, and few-shot fine-tuning. Top: unsupervised ( zero-shot ) results. Bottom: few-shot results using the best configuration (e.g, text+inst for emotion & topic).
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Dataset
Instruction
emotion
Identify the main emotion expressed in the given (summary of a | keyphrases of a) song text
genre
Identify the main genre of the given (summary of a | keyphrases of a) song text
stars
Identify the sentiment of the given (summary of an | keyphrases of an) Amazon review
product
Identify the category of an Amazon product based on the review (summary | keyphrases)
topic
Identify the topic or theme of the given (summary of a | keyphrases of a) news article
frame
Identify the framing of the given (summary of a | keyphrases of a) news article
Appendix
Table 2: Instructions provided to the embedding model. We adopted them to resemble those of the original authors.
Figure 3: Detailed evaluation with increasing number of labeled examples per class.
Dataset
Prompt
emotion-summary
Write a concise summary (maximum 5 sentences) that focuses on the emotional
tone of the following song lyrics. Analyze the lyrics to determine the main
emotion being conveyed and describe how it is expressed.
Conclude with an emotion categorization:
emotion-keyphrases
Generate keyphrases (max 5 phrases) that describe the emotional tone of
the following song lyrics. Focus on phrases that reflect the emotional tone,
Appendix
Table 3: Prompts used to rewrite the original documents. We use the same prompts for topics and frames.
Leveraging large language models (LLMs) to analyze complex documents -- such as academic papers, technical manuals, and financial reports -- has emerged as a mainstream and critical task in both research and industry. In practice, users must first filter relevant documents from large collections and then conduct in-depth analysis (e.g. question answering) over the selected subset, yet existing systems flatten documents into plain-text chunks, discarding the rich hierarchical structures (sections, tables, figures, equations) and degrading downstream performance. We present DocMaster, a hierarchical structure-aware document analysis system. DocMaster parses documents into hierarchical document trees preserving original layouts and constructs a structure-aware semantic index that enables accurate document filtering and in-depth analysis. We demonstrate DocMaster through an interactive web interface that enables users to upload document collections, construct tree-based and multi-view semantic indices, filter relevant documents via natural-language conditions, and perform follow-up question answering over the filtered results. The source code, data, and demo are available at https://doc-master.github.io/.
Ziqi Chen, Yingli Zhou, Fangyuan Zhang +3
The Chinese University of Hong Kong, Shenzhen · The Chinese University of Hong Kong · OceanBase, AntGroup
Recent text-clustering methods use large language models to propose a cluster taxonomy from a corpus and then assign each text to it. These pipelines are fundamentally programmatic: the sequence of LLM calls and the rules for stopping, merging, and splitting clusters are fixed in code in advance, so they generalise poorly across corpora of different structure and cannot easily incorporate user-supplied constraints such as a target cluster count or a clustering intent. We propose an agentic alternative in which an orchestrator LLM inspects the state of the discovery process at each step and dispatches one of a small set of specialised agents -- proposer, synthesizer, auditor, investigator, and critic -- adapting the pipeline to the corpus rather than executing a fixed one. On seven public text-clustering benchmarks the method achieves state-of-the-art performance, beating the strongest prior baseline by 47% in mean ARI.
ReadingMachine is a computational methodology for structured corpus reading that uses large language models to perform bounded reading operations over entire document collections. Rather than relying on retrieval or recursive summarization, the approach decomposes analysis into inspectable stages including insight extraction, semantic clustering, theme generation, and iterative omission detection. By delaying irreversible compression and explicitly tracking intermediate representations, the method prioritizes coverage, traceability, and preservation of disagreement across large corpora. The system is demonstrated on a heterogeneous corpus of 152 industrial policy documents, producing more than 17,500 extracted insights and a structured thematic map. ReadingMachine is released as an open-source experimental framework for large-scale qualitative synthesis and corpus analysis.