Chatting with Nature Journals Current Issue using a local Language Model • NatChat

📦 Overview

NatChat is an R package designed to summarize all papers in the current issues of journals published by the Nature Portfolio. It is inspired from the biorecap R package developed by the talented Stephen Turner. NatChat provides functions to:

Identify available Nature journals supported by the package.
Scrape and retrieve article abstracts from the latest issues of Nature journals.
Construct prompts to summarize articles using large language models (LLMs).
Generate natural language summaries via the ollama interface.
Format output for markdown tables, reports, or summaries.

This package is particularly useful for researchers, educators, and clinicians aiming to stay updated with the latest scientific literature across multiple disciplines through automated summarization and easy-to-use interfaces.

💻 Installation

You can install the development version of NatChat from GitHub with:

# Install `pak` package if not already installed
if (!requireNamespace("pak", quietly = TRUE)) {
  install.packages("pak")
}

# Install NatChat package
pak::pak("monahton/NatChat")

# Load the package
library(NatChat)

👉 Requirements: Ollama Setup

NatChat uses the ollamar package to interface with local large language models (LLMs) powered by Ollama. Before using NatChat’s summarization functions, ensure you have the Ollama software installed on your machine:

Download and install Ollama software from the official site.
Install the ollamar R package
Once installed, verify your setup in R:

library(NatChat)
library(ollamar)
check_ollama()    # Checks if Ollama is correctly connected

Make sure Ollama is running and accessible to R before proceeding with summarization tasks.

✅ Quick start

The easiest way to use NatChat is through the summarize_journal() function, which provides a one-command workflow for summarizing the articles in the current issue of a Nature journal and saving the results as both CSV and HTML reports.

This function handles the full NatChat pipeline: - Retrieves articles from the current issue of a specified journal, - Builds prompts for summarization, - Generates summaries using a local LLM via the Ollama interface, - Saves the results in well-formatted report files.

summarize_journal(journal = "Nature Biotechnology",
                  filename = "nature_biotechnology_summary",
                  outdir = ".",
                  model = "llama3.2",
                  save_csv = TRUE,
                  save_html = TRUE,
                  verbose = TRUE,
                  whitelist = NULL)

🧩 Arguments

Argument	Type	Description
`journal`	`character`	The name of the journal (e.g., `"Nature Medicine"`). Must match one of the options returned by `nat_journals()`.
`filename`	`character`	Base name for the output files (without extension). The function will append `.csv` and/or `.html` as appropriate.
`outdir`	`character`	Path to the directory where the output files will be saved. Default is the current directory (`"."`).
`model`	`character`	Name of the local language model to use via Ollama (e.g., `"llama3.1"`). Must be installed and available in the Ollama environment.
`save_csv`	`logical`	Whether to save the output summary as a CSV file. Default is `TRUE`.
`save_html`	`logical`	Whether to save the output summary as an HTML report. Default is `TRUE`.
`verbose`	`logical`	If `TRUE`, prints progress messages during execution. Useful for monitoring longer summarizations.
`whitelist`	`character` or `NULL`	Optional vector of keywords of interest to filter articles for summarization (e.g., “gene therapy”, “RNA-seq”). If `NULL`, all articles are processed.

📁 Functions Highlights

While the summarize_journal() function is the main user-facing tool in NatChat, designed to automate the end-to-end summarization workflow, internally, it builds upon a series of modular core functions listed below. These functions can also be used individually or customized for advanced usage and finer control over each step of the summarization process (eg., the prompt to summarize the article, number of sentences, …etc)

Function	Description
`nat_journals()`	Lists all supported Nature Portfolio journals available for scraping.
`get_articles()`	Retrieves article metadata and abstracts from the current issue.
`filter_articles()`	Filters articles based on user-defined criteria (e.g., keywords).
`build_prompt()`	Creates custom prompts from article titles and abstracts for LLM summarization.
`add_summary()`	Generates natural language summaries using the Ollama interface and LLMs.
`save_report()`	Saves the final formatted summaries into a report file.

🛠️ Development & Contributing

NatChat is an open-source project hosted on GitHub and is actively developed. Contributions and suggestions are welcome!

🔧 Open issues: https://github.com/monahton/NatChat/issues
📬 Email: aboualezz.monah@hsr.it
🤝 Contributions and Pull Requests encouraged!

✍️ Author

Monah Abou Alezz, PhD – aboualezz.monah@hsr.it.

San Raffaele Telethon Institute for Gene Therapy (SR-TIGET)
IRCCS San Raffaele Scientific Institute, Milan, Italy

🌍 Personal website

NatChat: Chatting with Nature Journals Current Issue using a local Language Model