Santosh-Gupta / Arxiv Manatee

This is the public repo we're going to be using to post updates on Arxiv-Manatee, which are going to be tools for searching through Arxiv-Sanity papers.

If you would like be involved, feel free to reach out to [email protected]. We could use those with experience with Tensorflow/Keras/Pytorch, NLP, abtractive summarization, and text/data processing, but we're open to considering anyone!

Our previous projects

https://github.com/re-search/DocProduct

https://github.com/re-search/gpt2-estimator

https://github.com/Santosh-Gupta/Research2Vec

https://github.com/Santosh-Gupta/Lit2Vec

Update 07-1-19

I am proud to present a summarization dataset for machine learning concepts.

https://drive.google.com/open?id=1B8qqHQNXZ4OVMKpGHAvrxPxdQlKt2Ia4

The dataset contains all 89252 machine learning papers ( cs.[CV|CL|LG|AI|NE]/stat.ML ) from Arxiv (1993-June 2019). The 'summary' is the title of the paper, and the 'source' is the abstract of each paper.

I am focused on concept summarization (as opposed to factual summarization), so all special characters and digits have been filtered out, so that the summarizars that are trained over this data can focus on the concepts described in the text.

6-7-19

Computer science summmarization datasets completed

This is a dataset of 5.6 million title / abstract data points, about 75% of which are from computer science papers (I tried my best to filter all non-CS papers (perhaps the non-CS papers add a bit of a "regularization" effect . . . ?) ) .

Title/Abstract pairs have been used to train biomedical summarizers [https://arxiv.org/pdf/1804.08875.pdf] , but I am doing a project on CS/ML papers so I made my own.

The dataset is basically a filtered version of the Semantic Scholar Corpus https://api.semanticscholar.org/corpus/

But it took some effort to produce it and I figure I may save some people time if they wanted the same.

This is a zip file containing 12 parquet files

https://drive.google.com/open?id=1WEdf-_au3vg2EzmWhawmW9xsYaHAE7iV

it's ~2.5 gb zipped, I think like 6 something gigs unzipped

This is the sqlite database version, 1 file

https://drive.google.com/open?id=1IhIaBD98BEseteAUi1S_f_SfIaUI8V4D

it's 2.5 gb zipped, 7.5 gb unzipped

To do

-Develop dataset (in progress)

-Figure out how to process data for training

https://github.com/yaserkl/RLSeq2Seq/tree/master/src/helper

-Train summarizer using this repo

https://github.com/yaserkl/RLSeq2Seq

-Download Arxiv papers from cs.[CV|CL|LG|AI|NE]/stat.ML

https://arxiv.org/help/bulk_data

Should be around ~73000 papers

The Arxiv-Sanity github repo may help figure out how to do this

https://github.com/karpathy/arxiv-sanity-preserver/blob/master/fetch_papers.py

-Find best solution to retieving on-disk text data using index or key.

Possible solution 1: sqlite

Possible solution 2. pickledb https://pythonhosted.org/pickleDB/ it looks like this is what arxiv-sanity uses https://github.com/karpathy/arxiv-sanity-preserver/blob/master/fetch_papers.py "The script is intended to enrich an existing database pickle (by default db.p)" from utils import Config, safe_pickle_dump # lets load the existing database to memory try: db = pickle.load(open(Config.db_path, 'rb'))

Possible solution 3. sqlitedict https://pypi.org/project/sqlitedict/

-Take text from extracted Arxiv papers and seperate them into sections/paragraph. Store each section/paragraph as a row in a row in a sqlite database. Should also contain Title, and arxiv link as seperate rows. Possibly abstract as well.

Possibly helpful resources

https://github.com/karpathy/arxiv-sanity-preserver/blob/master/parse_pdf_to_text.py

https://github.com/arxiv-vanity

https://github.com/arxiv-vanity/engrafo

-Figure out most efficient way to do embedding similarity search for large amounts of data.

Possible solutions 1 search through hdf5 data stored on disk

Possible solution 2 use Faiss IVF65536_HNSW32 index https://github.com/facebookresearch/faiss/wiki/Guidelines-to-choose-an-index