---
title: 'Project documentation template'
disqus: hackmd
---

Scientific Paper Search System (SPSS)
===

[TOC]

## About

The Scientific Papers Searching System allows users to apply full-text search across a limited set of scientific articles and search for methods that evaluate a specific set of data. The system is free to use.
With a continually expanded database of scientific papers, convenient in use query language and quick response from web service, SPSS will help users to produce more complete reviews.

## Getting started

If you are a total beginner to this, start here!

1. Visit https://kolchindru.github.io/
2. Start searching!

## User's Guide

### How to search by keyword?

#### Step 1
Actually searching SPSS is pretty easy. Just type what you’re interested in finding into the search box on the web site!


Press the button "ok".

#### Step 2

After you’ve searched, you’ll get top 20 related papers (paper names), sorted by their relevance. They are splitted to 4 pages so as not to overload the interface.
When user clicks on a paper, its pdf is previewed on the left on the site.


#### Step 3

You can choose different options of the view mode as well as download and print.


#### Step 4

You can repeat the search by clicking on the "magnifying glass" button in the right corner of the screen.


### How to search by dataset?

If a search query looks like ‘data-eval:”dataset1”, “dataset2”’ (no space after :), then our search engine tries to find articles which mention evaluation on any of these datasets.


If it doesn’t find any, it runs a standard search with ‘dataset1 dataset2’ query.
As a result 20 (or less if less dataset-evaluated relevant articles were found) articles are returned and displayed in the same way as with standard search.


## Admin's Guide

It is unavailable now due to the current selected deployment environment (heroku).

#### Step 1
First of all, in order to have access to the backend server via command line interface (as it was mentioned), an administrator should be logged in into heroku and be the application owner.

#### Step 2
Secondly, even if you are logged in, own the application and connect to the server via ssh connection, it is not granted that you will see the real application storage as it works in a fancy way (we will try to fix that later by using additional storage system like amazon storage service or another deployment system).

If the application is deployed on a local server manually directly or using Dockerfile, the administrator functionality (initializing database or adding new papers to database) can be invoked by running python scripts. The one for initializing parses all papers from https://proceedings.neurips.cc/ processes them and adds to a database, the one for adding new papers adds all the files listed in a specified directory to a database.