Mining Millions of Search Result Pages of Hundreds of Search Engines from 25 Years of Web Archives.
Mining Millions of Search Result Pages of Hundreds of Search Engines from 25 Years of Web Archives.
Start now by running your custom analysis/experiment, scraping your own query log, or just look at our example files.
The data in the Archive Query Log is highly sensitive (still, you
can re-crawl everything from the Wayback Machine). For that reason, we ensure that custom experiments or
analyses can not leak sensitive data (please get in touch if you have questions) by
using TIRA as a platform for custom analyses/experiments. In TIRA, you submit a Docker image that
implements your experiment. Your software is then executed in sandboxed mode (without internet connection) to ensure
that your software does not leak sensitive information. After your software execution finished, administrators will
review your submission and unblind it so that you can access the outputs.
Please refer to our dedicated TIRA tutorial as starting point for your experiments.
For running the CLI and crawl a query log on your own machine, please refer to the instructions for single-machine deployments. If instead you want to scale up and run the crawling pipelines on a cluster, please refer to the instructions for cluster deployments.
To run the Archive Query Log CLI on your machine, you can either use our PyPi package or the Docker image. (If you absolutely need to, you can also install the Python CLI or the Docker image from source.)
First you need to install Python 3.10 and pipx (this allows you to install the AQL CLI in a virtual environment). Then, you can install the Archive Query Log CLI by running:
pipx install archive-query-log
Now you can run the Archive Query Log CLI by running:
aql --help
First install Python 3.10, and clone this repository. From inside the repository directory, create a virtual environment and activate it:
python3.10 -m venv venv/
source venv/bin/activate
Install the Archive Query Log by running:
pip install -e .
Now you can run the Archive Query Log CLI by running:
aql --help
You only need to install Docker.
Note: The commands below use the syntax of the PyPi installation.
To run the same commands with the Docker installation, replace aql
with docker run -it -v "$(pwd)"/config.override.yml:/workspace/config.override.yml ghcr.io/webis-de/archive-query-log,
for example:
docker run -it -v "$(pwd)"/config.override.yml:/workspace/config.override.yml ghcr.io/webis-de/archive-query-log --help
First install Docker, and clone this repository. From inside the repository directory, build the Docker image like this:
docker build -t aql .
Note: The commands below use the syntax of the PyPi installation.
To run the same commands with the Docker installation, replace aql
with docker run -it -v "$(pwd)"/config.override.yml:/workspace/config.override.yml aql,
for example:
docker run -it -v "$(pwd)"/config.override.yml:/workspace/config.override.yml aql --help
Crawling the Archive Query Log requires access to an Elasticsearch cluster and some S3 block storage.
To configure access to the Elasticsearch cluster and S3,
add a config.override.yml file in the current directory, with the following contents.
Replace the placeholders with your actual credentials:
es:
host: "<HOST>"
port: 9200
username: "<USERNAME>"
password: "<PASSWORD>"
s3:
endpoint_url: "<URL>"
bucket_name: archive-query-log
access_key: "<KEY>"
secret_key: "<KEY>"
The crawling pipeline of the Archive Query Log can best be understood by looking at a small toy example. Here, we want to crawl and parse SERPs of the ChatNoir search engine from the Wayback Machine.
Add new web archive services (e.g., the Wayback Machine) to the AQL by running:
aql archives add
We maintain a list of compatible web archives below.
The web archives below are known to be compatible with the Archive Query Log crawler and can be used to mine SERPs.
| Name | CDX API URL | Memento API URL |
|---|---|---|
| Wayback Machine | https://web.archive.org/cdx/search/cdx | https://web.archive.org/web/ |
Add new search providers (e.g., Google) to the AQL by running:
aql providers add
A search provider can be any website that offers some kind of search functionality.
Ideally, you should also look at common prefixes of the URLs of the search results pages (e.g., /search for Google).
Narrowing down URL prefixes helps to not crawl too many captures that do not contain search results.
Refer to the import instructions below to import providers from the AQL-22 YAML file format.
Once you have added at least one archive and one search provider, we want to crawl archived captures of SERPs for each search provider and for each archive service, that is, we compute the cross product of archives and the search providers' domains and URL prefixes (roughly: archive×provider). Start building source pairs (i.e., archive–provider pairs) by running:
aql sources build
Running the command again after adding more archives or providers automatically creates the missing source pairs.
For each source pair, we now fetch captures from the archive service, that correspond to the provider's domain and URL prefix given in the source pair.
aql captures fetch
Again, running the command again after adding more source pairs automatically fetches the missing captures.
Not every capture necessarily points to a search engine result page (SERP). But usually, SERPs contain the user query in the URL, so we can filter out non-SERP captures by parsing the URLs.
aql serps parse url-query
Parsing the query from the capture URL will add SERPs to a new, more focused index that only contains SERPs. From the SERPs, we can also parse the page number and offset of the SERP, if available.
aql serps parse url-page
aql serps parse url-offset
All the above commands can be run in parallel, and they can be run multiple times to update the SERP index. Already parsed SERPs will be skipped.
Up to this point, we have only fetched the metadata of the captures, most prominently the URL. However, the snippets of the SERPs are not contained in the metadata, but only on the web page. So we need to download the actual web pages from the archive service.
aql serps download warc
This command will download the contents of each SERP to a WARC file that is stored in the configured S3 bucket. A pointer to the WARC file is stored in the SERP index so that we can quickly access a specific SERPs contents later.
We support automatically importing providers and parsers from the AQL-22 YAML-file format
(see data/selected-services.yaml).
To import the services and parsers from the AQL-22 YAML file, run:
aql providers import
aql parsers url-query import
aql parsers url-page import
aql parsers url-offset import
We also support importing a previous crawl of captures from the AQL-22 file system backend:
aql captures import aql-22
Running the Archive Query Log on a cluster is recommended for large-scale crawls. We provide a Helm chart that automatically starts crawling and parsing jobs for you and stores the results in an Elasticsearch cluster.
Just install Helm and configure kubectl for your cluster.
Crawling the Archive Query Log requires access to an Elasticsearch cluster and some S3 block storage.
Configure the Elasticsearch and S3 credentials in a values.override.yaml file like this:
elasticsearch:
host: "<HOST>"
port: 9200
username: "<USERNAME>"
password: "<PASSWORD>"
s3:
endpoint_url: "<URL>"
bucket_name: archive-query-log
access_key: "<KEY>"
secret_key: "<KEY>"
Let's deploy the Helm chart on the cluster (we're testing first with --dry-run to see if everything works):
helm upgrade --install --values helm/archive-query-log/values.override.yaml --dry-run archive-query-log helm/archive-query-log
If everything worked and the output looks good, you can remove the --dry-run flag to actually deploy the chart.
If you no longer need the chart, you can uninstall it:
helm uninstall archive-query-log
If you use the Archive Query Log dataset or the crawling code in your research, please cite the following paper describing the AQL and its use-cases:
Jan Heinrich Reimer, Sebastian Schmidt, Maik Fröbe, Lukas Gienapp, Harrisen Scells, Benno Stein, Matthias Hagen, and Martin Potthast. The Archive Query Log: Mining Millions of Search Result Pages of Hundreds of Search Engines from 25 Years of Web Archives. In Hsin-Hsi Chen et al., editors, 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2023), pages 2848–2860, July 2023. ACM.
You can use the following BibTeX entry for citation:
@InProceedings{reimer:2023,
author = {{Jan Heinrich} Reimer and Sebastian Schmidt and Maik Fr{\"o}be and Lukas Gienapp and Harrisen Scells and Benno Stein and Matthias Hagen and Martin Potthast},
booktitle = {46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2023)},
doi = {10.1145/3539618.3591890},
editor = {Hsin{-}Hsi Chen and Wei{-}Jou (Edward) Duh and Hen{-}Hsen Huang and Makoto P. Kato and Josiane Mothe and Barbara Poblete},
ids = {potthast:2023u},
isbn = {9781450394086},
month = jul,
numpages = 13,
pages = {2848--2860},
publisher = {ACM},
site = {Taipei, Taiwan},
title = {{The Archive Query Log: Mining Millions of Search Result Pages of Hundreds of Search Engines from 25 Years of Web Archives}},
url = {https://dl.acm.org/doi/10.1145/3539618.3591890},
year = 2023
}
Refer to the local Python installation instructions to set up the development environment and install the dependencies.
Then, also install the test dependencies:
pip install -e .[tests]
After having implemented a new feature, you should the check code format, inspect common LINT errors, and run all unit tests with the following commands:
flake8 archive_query_log # Code format
pylint archive_query_log # LINT errors
mypy archive_query_log # Static typing
bandit -c pyproject.toml -r archive_query_log # Security
pytest archive_query_log # Unit tests
At the moment, our workflow for adding new tests for parsers goes like this:
If you've found an important search provider to be missing from this query log, please suggest it by creating an issue. We also very gratefully accept pull requests for adding search providers or new parser configurations!
If you're unsure about anything, post an issue, or contact us:
We're happy to help!
This repository is released under the MIT license.
Files in the data/ directory are exempt from this license.
If you use the AQL in your research, we'd be glad if you'd cite us.
The Archive Query Log (AQL) is a previously unused, comprehensive query log collected at the Internet Archive over the last 25 years. Its first version includes 356 million queries, 166 million search result pages, and 1.7 billion search results across 550 search providers. Although many query logs have been studied in the literature, the search providers that own them generally do not publish their logs to protect user privacy and vital business data. Of the few query logs publicly available, none combines size, scope, and diversity. The AQL is the first to do so, enabling research on new retrieval models and (diachronic) search engine analyses. Provided in a privacy-preserving manner, it promotes open research as well as more transparency and accountability in the search industry.
FAQs
Mining Millions of Search Result Pages of Hundreds of Search Engines from 25 Years of Web Archives.
We found that archive-query-log demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 4 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Research
The Socket Research Team breaks down a malicious wrapper package that uses obfuscation to harvest credentials and exfiltrate sensitive data.
Research
Security News
Attackers used a malicious npm package typosquatting a popular ESLint plugin to steal sensitive data, execute commands, and exploit developer systems.
Security News
The Ultralytics' PyPI Package was compromised four times in one weekend through GitHub Actions cache poisoning and failure to rotate previously compromised API tokens.