CHRONOS: News Timeline Summarization Tool to Improve News Retrieval and Timeline Generation Efficiency

General Introduction

CHRONOS is a news timeline summarization tool developed by Alibaba NLP team. The tool generates timeline summaries of news events through iterative self-questioning.CHRONOS is not only capable of handling open-domain timeline summarization tasks, but also offers significant improvements in efficiency and scalability. Its unique dataset and algorithms allow it to excel in news retrieval and timeline generation for both researchers and news practitioners.

Demo address: https://modelscope.cn/studios/vickywu1022/CHRONOS

Function List

• News Timeline Generation: Generate a timeline of news events through iterative self-questioning.
• Open field timeline summary: Handles timeline summarization tasks in the open domain with a wide coverage of datasets.
• Efficient News Search: Utilize advanced search algorithms to quickly find relevant news.
• Dataset Release: Provide datasets summarized in the open field timeline for use in research.
• API integration: Supports API integration with Qwen, GPT, and other models for enhanced functionality.

Using Help

Installation process

1. Installation of dependencies: First, make sure that the required Python dependency packages are installed. Run the following command from the command line:

   pip install -r requirements.txt

2. Generate sample questions: Build a sample pool of topic questions for the dataset. Run the following command:

   python question_exampler.py

Or use the provideddata/question_examples.jsonfile, which contains sample questions for the Crisis, T17, and Open-TLS datasets.

3. Running CHRONOS: Execute the following command to complete the Open Field Timeline Summary task:

   python main.py --model_name "model name" --max_round "maximum number of iteration rounds" --dataset open --output "output directory" --question_exs

Guidelines for use

1. Replacing the API key: Before running thesrc/model.pyReplace the placeholders in your API key with your API key to invoke the Qwen or GPT model:

   DASHSCOPE_API_KEY = "Your API key"
   OPENAI_API_KEY = "Your API key"

At the same time, thesrc/searcher.pyReplace the Bing Web Search API key with your Bing Web Search API key:

   BING_SEARCH_KEY = "Your API key"

If you want CHRONOS to use the full page instead of just the snippet, add the following to thesrc/reader.pyin your JINA key:

   JINA_API_KEY = "Your API key"

2. Running Scripts: Run CHRONOS with the following command to experiment with the Open Domain Timeline Summary dataset:

   python main.py --model_name "model name" --max_round "maximum number of iteration rounds" --dataset open --output "output directory" --question_exs

Functional operation flow

1. News Timeline Generation: Through iterative self-questioning, CHRONOS is able to generate a detailed timeline of news events. Users can adjust the number of iteration rounds as needed to get a more comprehensive news summary.
2. Open field timeline summary: CHRONOS has a wide coverage of datasets that can handle a variety of open-domain timeline summarization tasks. Users can experiment with the provided datasets or with customized datasets.
3. Efficient News Search: CHRONOS utilizes advanced search algorithms to quickly find relevant news. Users can enhance the search function and improve efficiency through API integration.
4. Dataset Release: CHRONOS provides datasets summarizing open field timelines for use by researchers. Users can download and use these datasets for further research.
5. API integration: CHRONOS supports API integration with models such as Qwen, GPT, etc. Users can select the appropriate model to enhance the functionality as needed.

With the above steps, users can easily install and use CHRONOS to generate news timeline summaries and improve the efficiency of news retrieval and timeline generation.