Web Scraping all Questions from ResearchGate Search in Python

Intro

Currently, we don't have an API that supports extracting data from ResearchGate Search - Publications page.

This blog post is to show you a way you can do it yourself with the provided DIY solution below while we're working on releasing our proper API.

The solution can be used for personal use as it doesn't include the Legal US Shield that we offer for our paid production and above plans and has its limitations such as the need to bypass blocks, for example, CAPTCHA.

You can check our public roadmap to track the progress for this API:

What will be scraped

Prerequisites

Basic knowledge scraping with CSS selectors

If you haven't scraped with CSS selectors, there's a dedicated blog post of mine about how to use CSS selectors when web-scraping that covers what it is, pros and cons, and why they matter from a web-scraping perspective and show the most common approaches of using CSS selectors when web scraping.

Separate virtual environment

If you didn't work with a virtual environment before, have a look at the dedicated Python virtual environments tutorial using Virtualenv and Poetry blog post of mine to get familiar.

Reduce the chance of being blocked

There's a chance that a request might be blocked. Have a look at how to reduce the chance of being blocked while web-scraping, there are eleven methods to bypass blocks from most websites.

πŸ“ŒNote: This is not a strict requirement for this blog post.

Install libraries:

pip install parsel playwright

Full Code

from parsel import Selector
from playwright.sync_api import sync_playwright
import json


def scrape_researchgate_questions(query: str):
    with sync_playwright() as p:

        browser = p.chromium.launch(headless=True, slow_mo=50)
        page = browser.new_page(user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/101.0.4951.64 Safari/537.36")
        
        questions = []
        page_num = 1

        while True:
            page.goto(f"https://www.researchgate.net/search/question?q={query}&page={page_num}")
            selector = Selector(text=page.content())
            
            for question in selector.css(".nova-legacy-c-card__body--spacing-inherit"):
                title = question.css(".nova-legacy-v-question-item__title .nova-legacy-e-link--theme-bare::text").get().title()
                title_link = f'https://www.researchgate.net{question.css(".nova-legacy-v-question-item__title .nova-legacy-e-link--theme-bare::attr(href)").get()}'
                question_type = question.css(".nova-legacy-v-question-item__badge::text").get()
                question_date = question.css(".nova-legacy-v-question-item__meta-data-item:nth-child(1) span::text").get()
                snippet = question.css(".redraft-text::text").get()

                views = question.css(".nova-legacy-v-question-item__metrics-item:nth-child(1) .nova-legacy-e-link--theme-bare::text").get()
                views_link = question.css(".nova-legacy-v-question-item__metrics-item:nth-child(1) .nova-legacy-e-link--theme-bare::attr(href)").get()

                answer = question.css(".nova-legacy-v-question-item__metrics-item+ .nova-legacy-v-question-item__metrics-item .nova-legacy-e-link--theme-bare::text").get()
                answer_link = question.css(".nova-legacy-v-question-item__metrics-item+ .nova-legacy-v-question-item__metrics-item .nova-legacy-e-link--theme-bare::attr(href)").get()

                questions.append({
                    "title": title,
                    "link": title_link,
                    "snippet": snippet,
                    "question_type": question_type,
                    "question_date": question_date,
                    "views": {
                        "views_count": views,
                        "views_link": views_link
                        },
                    "answer": {
                        "answer_count": answer,
                        "answers_link": answer_link
                    }
                })

            print(f"page number: {page_num}")

            # checks if next page arrow key is greyed out `attr(rel)` (inactive) and breaks out of the loop
            if selector.css(".nova-legacy-c-button-group__item:nth-child(9) a::attr(rel)").get():
                break
            else:
                page_num += 1


        print(json.dumps(questions, indent=2, ensure_ascii=False))
        browser.close()
        

scrape_researchgate_questions(query="coffee")

Code explanation

Import libraries:

from parsel import Selector
from playwright.sync_api import sync_playwright
import json
Code Explanation
parsel to parse HTML/XML documents. Supports XPath.
playwright to render the page with a browser instance.
json to convert Python dictionary to JSON string.

Define a function and open a playwright with a context manager::

scrape_researchgate_questions(query="coffee"):
    with sync_playwright() as p:
        # ...
Code Explanation
query: str to tell Python that query should be an str.

Lunch a browser instance, open new_page with passed user-agent:

browser = p.chromium.launch(headless=True, slow_mo=50)
page = browser.new_page(user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/101.0.4951.64 Safari/537.36")
Code Explanation
p.chromium.launch() to launch Chromium browser instance.
headless to explicitly tell playwright to run in headless mode even though it's a defaut value.
slow_mo to tell playwright to slow down execution.
browser.new_page() to open new page. user_agent is used to act a real user makes a request from the browser. If not used, it will default to playwright value which is None. Check what's your user-agent.

Add a temporary list, set up a while loop, and open a new URL:

questions = []

while True:
    page.goto(f"https://www.researchgate.net/search/question?q={query}&page={page_num}")
    selector = Selector(text=page.content())
    # ...
Code Explanation
goto() to make a request to specific URL with passed query and page parameters.
Selector() to pass returned HTML data with page.content() and process it.

Iterate over author results on each page, extract the data and append to a temporary list:

for question in selector.css(".nova-legacy-c-card__body--spacing-inherit"):
    title = question.css(".nova-legacy-v-question-item__title .nova-legacy-e-link--theme-bare::text").get().title()
    title_link = f'https://www.researchgate.net{question.css(".nova-legacy-v-question-item__title .nova-legacy-e-link--theme-bare::attr(href)").get()}'
    question_type = question.css(".nova-legacy-v-question-item__badge::text").get()
    question_date = question.css(".nova-legacy-v-question-item__meta-data-item:nth-child(1) span::text").get()
    snippet = question.css(".redraft-text::text").get()

    views = question.css(".nova-legacy-v-question-item__metrics-item:nth-child(1) .nova-legacy-e-link--theme-bare::text").get()
    views_link = question.css(".nova-legacy-v-question-item__metrics-item:nth-child(1) .nova-legacy-e-link--theme-bare::attr(href)").get()

    answer = question.css(".nova-legacy-v-question-item__metrics-item+ .nova-legacy-v-question-item__metrics-item .nova-legacy-e-link--theme-bare::text").get()
    answer_link = question.css(".nova-legacy-v-question-item__metrics-item+ .nova-legacy-v-question-item__metrics-item .nova-legacy-e-link--theme-bare::attr(href)").get()

    questions.append({
        "title": title,
        "link": title_link,
        "snippet": snippet,
        "question_type": question_type,
        "question_date": question_date,
        "views": {
            "views_count": views,
            "views_link": views_link
            },
        "answer": {
            "answer_count": answer,
            "answers_link": answer_link
        }
    })

Code Explanation
css() to parse data from the passed CSS selector(s). Every CSS query traslates to XPath using csselect package under the hood.
::text/::attr(attribute) to extract textual or attribute data from the node.
get()/getall() to get actual data from a matched node, or to get a list of matched data from nodes.
xpath("normalize-space()") to parse blank text node as well. By default, blank text node is be skipped by XPath.

Check if the next page is present and paginate:

# checks if the next page arrow key is greyed out `attr(rel)` (inactive) -> breaks out of the loop
if selector.css(".nova-legacy-c-button-group__item:nth-child(9) a::attr(rel)").get():
    break
else:
    page_num += 1

Print extracted data, and close browser instance:

print(json.dumps(publications, indent=2, ensure_ascii=False))

browser.close()

# call the function
scrape_researchgate_questions(query="coffee")

Part of the JSON output:

[
  {
    "title": "Any Recommendations On An Inexpensive Coffee Grinder To Grind Up Bark Samples To Measure Ph?",
    "link": "https://www.researchgate.netpost/Any_recommendations_on_an_inexpensive_coffee_grinder_to_grind_up_bark_samples_to_measure_pH?_sg=tsmZvLsXrFpn6TG77ljxS8pVJhdOMYVlqqYhQl0BszqPCDW1__lnpczwZl8XJiVROJ8_8G8jaerzpX8",
    "snippet": "We are folloiwng protocol by Hansen et al. (2015) Sci. Pharm. They recommend a Rancilio coffee grinder but these are several hundred dollars. Hoping to use something a little less expensive.",
    "question_type": "Question",
    "question_date": "Oct 2017",
    "views": {
      "views_count": "97 Views",
      "views_link": "post/Any_recommendations_on_an_inexpensive_coffee_grinder_to_grind_up_bark_samples_to_measure_pH?_sg=tsmZvLsXrFpn6TG77ljxS8pVJhdOMYVlqqYhQl0BszqPCDW1__lnpczwZl8XJiVROJ8_8G8jaerzpX8"
    },
    "answer": {
      "answer_count": "2 Answers",
      "answers_link": "https://www.researchgate.netpost/Any_recommendations_on_an_inexpensive_coffee_grinder_to_grind_up_bark_samples_to_measure_pH?_sg=tsmZvLsXrFpn6TG77ljxS8pVJhdOMYVlqqYhQl0BszqPCDW1__lnpczwZl8XJiVROJ8_8G8jaerzpX8"
    }
  }, ... other questions
  {
    "title": "Are There Any Ways To Find The Concentration Of A Solution Where Its Chemical Formula And Number Of Moles Are Unknown? ",
    "link": "https://www.researchgate.netpost/Are_there_any_ways_to_find_the_concentration_of_a_solution_where_its_chemical_formula_and_number_of_moles_are_unknown?_sg=6W-hvIYx-FRel_YiWd62lbksTzeWP7GVkZ3tVO6SgZI7F_czhLz_oFCduq9DVhrhvIUy97168wXrn30",
    "snippet": "A comprehensive way to find the concentration of random solutions would enhance benefits related with health, industry, technology and commercial aspects. Although beer lambert law is a solution, there are some cases where Epsilon is unknown (Example: A Coca-Cola drink or a cup of coffee). In this cases, proper a​l​t​",
    "question_type": "Question",
    "question_date": "Jan 2022",
    "views": {
      "views_count": "742 Views",
      "views_link": "post/Are_there_any_ways_to_find_the_concentration_of_a_solution_where_its_chemical_formula_and_number_of_moles_are_unknown?_sg=6W-hvIYx-FRel_YiWd62lbksTzeWP7GVkZ3tVO6SgZI7F_czhLz_oFCduq9DVhrhvIUy97168wXrn30"
    },
    "answer": {
      "answer_count": "4 Answers",
      "answers_link": "https://www.researchgate.netpost/Are_there_any_ways_to_find_the_concentration_of_a_solution_where_its_chemical_formula_and_number_of_moles_are_unknown?_sg=6W-hvIYx-FRel_YiWd62lbksTzeWP7GVkZ3tVO6SgZI7F_czhLz_oFCduq9DVhrhvIUy97168wXrn30"
    }
  }
]

Join us on Twitter | YouTube

Add a Feature RequestπŸ’« or a Bug🐞