Institutional Books - Enriched Text: A customizable multilingual open-source pipeline for denoising, deduplicating, and annotating OCR text at scale

Institutional Books - Visual Elements: An open-source pipeline for extracting, classifying, deduplicating, and captioning visual elements from digital book collections

Historical book collections contain rich visual elements - such as illustrations, photographs, engravings, and decorative art - that are frequently under-explored in large-scale digitization projects. While Optical Character Recognition (OCR) has standardized the extraction of textual content, these visual components offer a layer of nuance and context that remains largely untapped by automated text extraction workflows. This technical report introduces Institutional Books - Visual Elements, an open-source end-to-end pipeline for detecting, classifying, deduplicating, and captioning visual elements from historical book collections. Alongside this pipeline, we release an initial dataset of 22.6 million visual elements extracted from the 983,004 scanned volumes that comprise the Institutional Books: Harvard Library dataset. This work contributes to ongoing, community-wide efforts to enable new use cases for digitized library collections through computational access, from artificial intelligence model training to digital humanities research.

A Reassessment of TextRank: Graph-Based Extractive Summarization in the Era of Large Language Models- A Systematic Review

Despite the emergence of Large Language Models (LLMs), graph-based extractive summarization and specifically the TextRank algorithm have retained architectural significance in high-stakes domains including legal, medical, scientific, and journalistic text processing. This survey presents a systematic review of 121 references spanning 2002 to 2026, tracing the evolution of TextRank-based approaches into hybrid LLM pipelines and advancing three qualified arguments.First, TextRank provides architectural advantages that persist even as LLM context windows grow: source-level provenance that grounded generation methods only partially replicate; approximately 75% token consumption reduction at any fixed per-token rate; and deterministic, auditable extraction consistent with emerging regulatory transparency requirements in legal and clinical domains.Second, LLMs exhibit hallucination rates of 15–30% on domain-specific text and produce no source audit trail; TextRank pre-filtering substantially mitigates both limitations, while grounded generation alternatives address them at the generation layer. Third, hybrid TextRank→LLM pipelines achieve ROUGE-1 = 0.41 on CNN/DailyMail and 0.54–0.76 on domain-specific benchmarks; cross-benchmark ROUGE comparisons are not valid across different benchmark types and should not be interpreted as a unified performance ranking. The performance comparisons are derived from the reviewed literature; the layered pipeline architecture, limitations taxonomy, and research agenda are original contributions of this survey."

Beyond Archives: Designing a Segmentation and ATR Workflow for Mid-Twentieth-Century Italian Typescripts in a Relational Digital Library

Automated Summarization Tool

As the size of file is increasing in non-structured text format in academic, corporate, law and organisation, there is a severe need of information-extraction and text-summarization systems. Doing summary manually is a time consuming, inconsistent and unscalable process. This paper presents the design realization and evaluation of an Automated Summarization Tool (AST) which is a document intelligence platform based on google gemini 2.5 flash. The platform employs map-reduce summarization for long documents, the use of SHA-256 hash for caching, a RAG-lite chat module grounded in the source document enabling conversational chats, role- and toneadaptive prompt engineering, and a JSON-based output schema coupling every extracted key point with a verbatim quote and location from the source for traceability, thereby preventing unnecessary API calls. The complete system features a Gradio web interface. It is deployed as a zero-infrastructure Google Colab notebook. Thus, no dedicated server or installation is required. In internal, multi-domain test corpora, our platform achieves a ROUGE-1 score of 58.17. This is the highest we obtain out of six different summarization systems with which we compare against other methods. And it outperforms the strongest fine-tuned transformer baselines (PEGASUS, BART) by ~14 points. And it is clearly ahead of BERTSUM-ext (a strong transformer baseline), Pointer-Generator Network, TextRank.

Natural Language Code Retrieval for 1C:Enterprise: An Open Benchmark and Efficient Bi-Encoder

Natural language code retrieval is a rapidly evolving task in computer science. However, the 1C:Enterprise ecosystem combines Russian syntax with highly domain-specific terminology, for which open datasets and specialized models have been virtually non-existent. We present a comprehensive pipeline for 1C code retrieval: an open benchmark of 3,413 real-world, PII-scrubbed query-code pairs, a reproducible evaluation harness, and a specialized bi-encoder. To overcome scarce labeled data, we fine-tune on 784,057 synthetic triplets generated by google/gemma-4-26B-A4B-it from public code repositories, using Matryoshka Representation Learning (MRL) and a privacy-aware tokenizer. Because the benchmark subsets differ in size, we report balanced-subset macro, query-weighted micro, and forum-only results. Our model reaches 0.5992 balanced macro nDCG@10, 0.5044 micro, and 0.4617 on forum, versus 0.4932 macro for the baseline architecture and 0.5404 for google/embeddinggemma-300m. Removing every benchmark example flagged by the conservative exact/13-gram overlap audit leaves 0.6011 balanced macro (0.5010 micro), indicating that detected train-benchmark overlap does not explain the headline result. MRL truncation to 256 dimensions preserves 99.9% of retrieval quality while reducing dense-index storage and exact similarity arithmetic by a factor of three.

MORE: A Multilingual Document Parsing Benchmark and Evaluation

Multilingual documents encapsulate rich regional cultures, scientific discoveries, and historical records. Parsing this content into structured, machine-readable formats is critical for unlocking global knowledge. However, existing benchmarks predominantly focus on high-resource languages like English and Chinese, creating an evaluation blind spot concerning model performance on other languages. While recent Vision-Language Models (VLMs) claim support for hundreds of languages, the lack of ground truth makes it impossible to empirically verify these capabilities. To bridge this gap, we introduce MORE, a large-scale benchmark designed for multilingual document parsing evaluation. MORE distinguishes itself through three key dimensions: (1) Unprecedented Scale: It covers 149 languages, making it the most linguistically diverse benchmark to date; (2) Structural Complexity: Unlike previous works, it extends evaluation beyond plain text to include structural elements such as code blocks, tables, and catalogs; and (3) Data Authenticity: All samples are curated from real-world documents via a model-assisted, human-refined annotation pipeline. We evaluate state-of-the-art models using MORE, establishing new performance baselines for long-tail languages and validating the benchmark's effectiveness in diagnosing model capabilities in realistic, diverse scenarios. The MORE dataset will be available at https://github.com/zimoqingfeng/MORE.