BenJamesbabala/AI-Paper-Collector

Fully-automated scripts for collecting AI-related papers

AI-Paper-Collector

Web demo: https://ai-paper-collector.vercel.app/ (recommended)

Colab notebook: here

Motivation

Fully-automated scripts for collecting AI-related papers. Support fuzzy and exact search for paper titles.

Search Categories

- [EMNLP 2019-2021] [ACL 2019-2022] [NAACL 2019-2022] [COLING 2020-2022] 
- [ICASSP 2019-2022] [WWW 2019-2022] [ICLR 2019-2022] [ICML 2019-2022] 
- [AAAI 2019-2022] [IJCAI 2019-2022] [CVPR 2019-2022] [ICCV 2019-2021] 
- [MM 2019-2022] [KDD 2019-2022] [CIKM 2019-2022] [SIGIR 2019-2022] 
- [WSDM 2019-2022] [ECIR 2019-2022] [ECCV 2020-2022] [COLT 2019-2022] 
- [AISTATS 2019-2022] [INTERSPEECH 2019-2022] [ISWC 2019-2022] [JMLR 2019-2022] 
- [VLDB 2019-2021] [ICME 2019-2022] [TIP 2020-2022] [TPAMI 2020-2022] 
- [RECSYS 2019-2022] [TKDE 2020-2022] [TOIS 2020-2022] [ICDM 2019-2021] 
- [TASLP 2020-2022] [BMVC 2019-2021] [MICCAI 2019-2022] [NIPS 2019-2022] 
- [MLSYS 2020-2022] [WACV 2020-2022] 

Installation

Current installation is to clone this repo.

git clone https://github.com/MLNLP-World/AI-Paper-Collector.git
cd AI-Paper-Collector
pip install -r requirements.txt

Usage(v0.1.0)

We provide three usage modes, the first is interactive (main.py), the second is command-line (cli_main.py) and the other is web interface (app.py). The interactive mode is recommended for the first time users.

Interactive Usage with Example

To start the interactive, type:

python main.py

The initialization of indices and posting-list may take several seconds when you first use this script, after that they will be stored in ./output/ and you don't have to initialized them again. When the ./cache/cache.json is updated, you should manually delete ./output/postings.pkl and run main.py again.

Serveral steps to interactively search paper.

1. the keyword query
2. search mode (exact, fuzzy or boolean)
3. (fuzzy) threshold
4. the limit of results
5. a list of conferences, separated by comma
6. the file path of the output (top-5 for command preview, all results in this file)

E.g.

[+] Initializing System...
[+] Loading from cache...
[+] Enter your query: few-shot

[+] Select search mode:
	[1] Exact
	[2] Fuzzy
  [3] Boolean
[+] Enter a number between 1 to 3: 2
[+] Enter threshold between 0 and 100 (default: 50):
[+] Enter limit >= 0 (default: None):
[+] Enter the list of confs separated by comma
	E.g. "ACL,CVPR" or "AAAI" or enter nothing for all confs
[+] Enter your list of conferences (default: All Confs): SIGIR,WSDM,CIKM

[+] Search Results:
[=] Only show Top-5, Please Save results to see all.
[1] [CIKM2021] REFORM: Error-Aware Few-Shot Knowledge Graph Completion.
[2] [CIKM2021] Boosting Few-shot Abstractive Summarization with Auxiliary Tasks.
[3] [CIKM2021] Multi-objective Few-shot Learning for Fair Classification.
[4] [CIKM2020] Graph Few-shot Learning with Attribute Matching.
[5] [CIKM2020] Few-shot Insider Threat Detection.

[+] Enter Save filename:
[+] Writing results to output/fuzzy_None_SIGIR_WSDM_CIKM_few-shot.txt
[+] Writing results Done!

Boolean Query Rules:

For boolean search, you can use the standard boolean expressions with [AND, OR, NOT] and brackets. For example, you can write your queries like:

1. language AND generation AND (pre-train OR pretrain)
2. (dialogue OR dialog) AND generation AND NOT (response AND selection)
3. toxic AND (dialogue OR conversation OR dialog)

Note that when you want to search for a phrase (e.g. contrastive learning), you should type contrastive AND learning instead of leaving blank between the words like contrastive learning.

The boolean query allows you to search exactly the key-words that you are interested in. Besides, it also helps to include the near-synonyms (like dialog, dialogue and conversation) and exclude the words that you are not interested in (like the second example).

Command-line Usage

For command-line usage, you can use the following commands:

# -q, --query:     the input query, and the content with multiple words should be wrapped in quotation marks
# -m, --mode:      the search mode: fuzzy or exact, default is exact
# -t, --threshold: the threshold for the fuzzy search, default is 50
# -l, --limit:     the limit num of the fuzzy search result, default is None
# -c, --conf:      the list of the conferences needs to search, default is all
# -o, --output:    the output file name, default is [mode]_[threshold]_[confs]_[query].txt
# -f, --force:     force to update the cache file incrementally
python cli_main.py --query QUERY \
    [--mode {fuzzy,exact}] \
    [--threshold THRESHOLD] [--limit LIMIT] [--conf CONF] \
    [--output OUTPUT] [--force]

E.g.

# Note that the input query must be enclosed in `""`, such as "few shot".
python cli_main.py -q "few shot" -m fuzzy -l 10 -t 10 -c AAAI,ACL -o results.txt

Web interface Usage

For web interface usage, you can use the following commands:

pip install -r requirements.txt
python app.py

Then open the following URL: http://localhost:5000

E.g.

How to add new conferences from DBLP

Automatically Updating via an issue-triggered workflow

If anyone wants to add a new list of conferences. please raise an issue following the format of this one. We will check and label it, then the workflow will run automatically. issue format

For users who clone the project to use

• add new conferences by modifying the conf/dblp_conf.json file

[
    # add the name and dblp_url of the new conf
    {
        "name": "WWW2021",
        "url": "https://dblp.org/db/conf/www/www2021.html"
    },
    ...
]

• run the script

# force to update the cache file incrementally
python cli_main.py --query '' --force

Disclaimer

Since the tool is in the development stage, we can not guarantee that the papers found will meet your needs. I hope for your understanding. In addition, all the results come from DBLP, ACL, NIPS, OpenReview, if this violates your copyright, you can contact us at any time, we will delete it as soon as possible, thank you:)

Organizers

Contributors

Thanks to the contributors: