Overview

Teaching: 15 min
Exercises: 0 min

Questions

• What is web scraping and why is it useful?

• What are typical use cases for web scraping?

• When is web scraping OK and when is it not?

• Is web scraping legal? Can I get into trouble?

• How can I make sure I’m doing the right thing?

• What can I do with the data that I’ve scraped?

Objectives

• Introduce the concept of structured data

• Discuss how data can be extracted from web pages

• Introduce the examples that will be used in this lesson

• Discuss the legal implications of web scraping

• Establish a code of conduct

What is web scraping?

Web scraping is a technique for extracting information from websites. This can be done manually but it is usually faster, more efficient and less error-prone to automate the task.

Web scraping allows you to acquire non-tabular or poorly structured data from websites and convert it into a usable, structured format, such as a .csv file or spreadsheet.

Scraping is about more than just acquiring data: it can also help you archive data and track changes to data online.

It is closely related to the practice of web indexing, which is what search engines like Google do when mass-analysing the Web to build their indices. But contrary to web indexing, which typically parses the entire content of a web page to make it searchable, web scraping targets specific information on the pages visited.

For example, online stores will often scour the publicly available pages of their competitors, scrape item prices, and then use this information to adjust their own prices. Another common practice is “contact scraping” in which personal information like email addresses or phone numbers is collected for marketing purposes.

Web scraping is also increasingly being used by scholars to create data sets for text mining projects; these might be collections of journal articles or digitised texts. The practice of data journalism, in particular, relies on the ability of investigative journalists to harvest data that is not always presented or published in a form that allows analysis.

Before you get started

As useful as scraping is, there might be better options for the task. Choose the right (i.e. the easiest) tool for the job.

• Check whether or not you can easily copy and paste data from a site into Excel or Google Sheets. This might be quicker than scraping.
• Check if the site or service already provides an API to extract structured data. If it does, that will be a much more efficient and effective pathway. Good examples are the Facebook API, the Twitter APIs or the YouTube comments API.
• For much larger needs, Freedom of information requests can be useful. Be specific about the formats required for the data you want.

we will be discussing some of the issues to be aware of when scraping websites, and we will establish a code of conduct (below) to guide our web scraping projects.

This section does not constitute legal advice

Please note that the information provided on this page is for information purposes only and does not constitute professional legal advice on the practice of web scraping.

If you are concerned about the legal implications of using web scraping on a project you are working on, it is probably a good idea to seek advice from a professional, preferably someone who has knowledge of the intellectual property (copyright) legislation in effect in your country.

Don’t break the web: Denial of Service attacks

The first and most important thing to be careful about when writing a web scraper is that it typically involves querying a website repeatedly and accessing a potentially large number of pages. For each of these pages, a request will be sent to the web server that is hosting the site, and the server will have to process the request and send a response back to the computer that is running our code. Each of these requests will consume resources on the server, during which it will not be doing something else, like for example responding to someone else trying to access the same site.

If we send too many such requests over a short span of time, we can prevent other “normal” users from accessing the site during that time, or even cause the server to run out of resources and crash.

In fact, this is such an efficient way to disrupt a web site that hackers are often doing it on purpose. This is called a Denial of Service (DoS) attack.

Since DoS attacks are unfortunately a common occurence on the Internet, modern web servers include measures to ward off such illegitimate use of their resources. They are watchful for large amounts of requests appearing to come from a single computer or IP address, and their first line of defense often involves refusing any further requests coming from this IP address.

A web scraper, even one with legitimate purposes and no intent to bring a website down, can exhibit similar behaviour and, if we are not careful, result in our computer being banned from accessing a website.

The good news is that a good web scraper, such as Scrapy, recognizes that this is a risk and includes measures to prevent our code from appearing to launch a DoS attack on a website. This is mostly done by inserting a random delay between individual requests, which gives the target server enough time to handle requests from other users between ours.

This is Scrapy’s default behaviour, and it should prevent most scraping projects from ever causing problems. To be on the safe side, however, it is good practice to limit the number of pages we are scraping while we are still writing and debugging our code. This is why in the previous section, we imposed a limit of five pages to be scraped, which we only removed when we were reasonably certain the scraper was working as it should.

Limiting requests to a particular domain, by using Scrapy’s allowed_domains property is another way to make sure our code is not going to start scraping the entire Internet by mistake.

Thanks to the defenses web servers use to protect themselves against DoS attacks and Scrapy’s measure to avoid inadvertently launching such an attack, the risks of causing trouble is limited.

Don’t steal: Copyright and fair use

It is important to recognize that in certain circumstances web scraping can be illegal. If the terms and conditions of the web site we are scraping specifically prohibit downloading and copying its content, then we could be in trouble for scraping it.

In practice, however, web scraping is a tolerated practice, provided reasonable care is taken not to disrupt the “regular” use of a web site, as we have seen above.

In a sense, web scraping is no different than using a web browser to visit a web page, in that it amounts to using computer software (a browser vs a scraper) to acccess data that is publicly available on the web.

In general, if data is publicly available (the content that is being scraped is not behind a password-protected authentication system), then it is OK to scrape it, provided we don’t break the web site doing so. What is potentially problematic is if the scraped data will be shared further. For example, downloading content off one website and posting it on another website (as our own), unless explicitely permitted, would constitute copyright violation and be illegal.

However, most copyright legislations recognize cases in which reusing some, possibly copyrighted, information in an aggregate or derivative format is considered “fair use”. In general, unless the intent is to pass off data as our own, copy it word for word or trying to make money out of it, reusing publicly available content scraped off the internet is OK.

Better be safe than sorry

Be aware that copyright and data privacy legislation typically differs from country to country. Be sure to check the laws that apply in your context. For example, in Australia, it can be illegal to scrape and store personal information such as names, phone numbers and email addresses, even if they are publicly available.

If you are looking to scrape data for your own personal use, then the above guidelines should probably be all that you need to worry about. However, if you plan to start harvesting a large amount of data for research or commercial purposes, you should probably seek legal advice first.

If you work in a university, chances are it has a copyright office that will help you sort out the legal aspects of your project. The university library is often the best place to start looking for help on copyright.

Be nice: ask and share

Depending on the scope of your project, it might be worthwhile to consider asking the owners or curators of the data you are planning to scrape if they have it already available in a structured format that could suit your project. If your aim is do use their data for research, or to use it in a way that could potentially interest them, not only it could save you the trouble of writing a web scraper, but it could also help clarify straight away what you can and cannot do with the data.

On the other hand, when you are publishing your own data, as part of a research project, documentation or a public website, you might want to think about whether someone might be interested in getting your data for their own project. If you can, try to provide others with a way to download your raw data in a structured format, and thus save them the trouble to try and scrape your own pages!

Web scraping code of conduct

This all being said, if you adhere to the following simple rules, you will probably be fine.

1. Ask nicely. If your project requires data from a particular organisation, for example, you can try asking them directly if they could provide you what you are looking for. With some luck, they will have the primary data that they used on their website in a structured format, saving you the trouble.
2. Don’t download copies of documents that are clearly not public. For example, academic journal publishers often have very strict rules about what you can and what you cannot do with their databases. Mass downloading article PDFs is probably prohibited and can put you (or at the very least your friendly university librarian) in trouble. If your project requires local copies of documents (e.g. for text mining projects), special agreements can be reached with the publisher. The library is a good place to start investigating something like that.
3. Check your local legislation. For example, certain countries have laws protecting personal information such as email addresses and phone numbers. Scraping such information, even from publicly avaialable web sites, can be illegal (e.g. in Australia).
4. Don’t share downloaded content illegally. Scraping for personal purposes is usually OK, even if it is copyrighted information, as it could fall under the fair use provision of the intellectual property legislation. However, sharing data for which you don’t hold the right to share is illegal.
5. Share what you can. If the data you scraped is in the public domain or you got permission to share it, then put it out there for other people to reuse it (e.g. on datahub.io). If you wrote a web scraper to access it, share its code (e.g. on GitHub) so that others can benefit from it.
6. Don’t break the Internet. Not all web sites are designed to withstand thousands of requests per second. If you are writing a recursive scraper (i.e. that follows hyperlinks), test it on a smaller dataset first to make sure it does what it is supposed to do. Adjust the settings of your scraper to allow for a delay between requests. By default, Scrapy uses conservative settings that should minimize this risk.
7. Publish your own data in a reusable way. Don’t force others to write their own scrapers to get at your data. Use open and software-agnostic formats (e.g. JSON, XML), provide metadata (data about your data: where it came from, what it represents, how to use it, etc.) and make sure it can be indexed by search engines so that people can find it.

Going further

This lesson only provides an introduction to the practice of web scraping and highlights some of the tools available. Scrapy has many more features than those mentioned in the previous section, be sure to refer to its full documentation for details.

Example: scraping government websites for contact addresses

In this lesson, we will extract contact information from government websites that list the members of various constituencies. Librarians could use this example to scrape information from any site listing contact details.

Let’s start by looking at the current list of members of the Canadian parliament, which is available on the Parliament of Canada website.

This is how this page appears in November 2016:

There are several features (circled in the image above) that make the data on this page easier to work with. The search, reorder, refine features and display modes hint that the data is actually stored in a (structured) database before being displayed on this page. The data can be readily downloaded either as a comma separated values (.csv) file or as XML for re-use in their own database, spreadsheet or computer program.

Even though the information displayed in the view above is not labelled, anyone visiting this site with some knowledge of Canadian geography and politics can see what information pertains to the politicians’ names, the geographical area they come from and the political party they represent. This is because human beings are good at using context and prior knowledge to quickly categorise information.

Computers, on the other hand, cannot do this unless we provide them with more information. Fortunately, if we examine the source HTML code of this page, we can see that the information displayed is actually organised inside labelled elements:

(...)
<div>
    <a href="/Parliamentarians/en/members/Ziad-Aboultaif(89156)">
        <img alt="Photo - Ziad Aboultaif - Click to open the Member of Parliament profile" title="Photo - Ziad Aboultaif - Click to open the Member of Parliament profile" src="http://www.parl.gc.ca/Parliamentarians/Images/OfficialMPPhotos/42/AboultaifZiad_CPC.jpg" class="picture" />
        <div class="full-name">
		    <span class="honorific"><abbr></abbr></span>
            <span class="first-name">Ziad</span>
            <span class="last-name">Aboultaif</span>
        </div>
    </a>
    <div class="caucus-banner" style="background-color:#002395"></div>
    <div class="caucus">Conservative</div>
    <div class="constituency">Edmonton Manning</div>
    <div class="province">Alberta</div>        
</div>
(...)

Thanks to these labels, we could relatively easily instruct a computer to look for all parliamentarians from Alberta and list their names and caucus information.

Structured vs unstructured data

When presented with information, human beings are good at quickly categorizing it and extracting the data that they are interested in. For example, when we look at a magazine rack, provided the titles are written in a script that we are able to read, we can rapidly figure out the titles of the magazines, the stories they contain, the language they are written in, etc. and we can probably also easily organize them by topic, recognize those that are aimed at children, or even whether they lean toward a particular end of the political spectrum. Computers have a much harder time making sense of such unstructured data unless we specifically tell them what elements data is made of, for example by adding labels such as this is the title of this magazine or this is a magazine about food. Data in which individual elements are separated and labelled is said to be structured.

Let’s look now at the current list of members for the UK House of Commons.

This page also displays a list of names, political and geographical affiliation. There is a search box and a filter option, but no obvious way to download this information and reuse it.

Here is the code for this page:

(...)
<table>
    <tbody>
        (...)
        <tr id="ctl00_ctl00_(...)_trItemRow" class="first">
            <td>Aberavon</td>
            <td id="ctl00_ctl00_(...)_tdNameCellRight">
                <a id="ctl00_ctl00_(...)_hypName" href="http://www.parliament.uk/biographies/commons/stephen-kinnock/4359">Kinnock, Stephen</a>(Labour)
            </td>
        </tr>
        (...)
    </tbody>
</table>
(...)

We see that this data has been structured for displaying purposes (it is arranged in rows inside a table) but the different elements of information are not clearly labelled.

What if we wanted to download this dataset and, for example, compare it with the Canadian list of MPs to analyze gender representation, or the representation of political forces in the two groups? We could try copy-pasting the entire table into a spreadsheet or even manually copy-pasting the names and parties in another document, but this can quickly become impractical when faced with a large set of data. What if we wanted to collect this information for every country that has a parliamentary system?

Fortunately, there are tools to automate at least part of the process. This technique is called web scraping.

“Web scraping (web harvesting or web data extraction) is a computer software technique of extracting information from websites.” (Source: Wikipedia)

Web scraping typically targets one web site at a time to extract unstructured information and put it in a structured form for reuse.

In this lesson, we will continue exploring the examples above and try different techniques to extract the information they contain. But before we launch into web scraping proper, we need to look a bit closer at how information is organized within an HTML document and how to build queries to access a specific subset of that information.

References

• Web Scraping (Wikipedia)
• The Data Journalism Handbook: Getting Data from the Web
• The Web scraping Wikipedia page has a concise definition of many concepts discussed here.
• The School of Data Handbook has a short introduction to web scraping, with links to resources e.g. for data journalists.
• This blog has a discussion on the legal aspects of web scraping.
• Scrapy documentation
• morph.io is a cloud-based web scraping platform that supports multiple frameworks, interacts with GitHub and provides a built-in way to save and share extracted data.
• import.io is a commercial web-based scraping service that requires little coding.
• Software Carpentry is a non-profit organisation that runs learn-to-code workshops worldwide. All lessons are publicly available and can be followed indepentently. This lesson is heavily inspired by Software Carpentry.
• Data Carpentry is a sister organisation of Software Carpentry focused on the fundamental data management skills required to conduct research.
• Library Carpentry is another Software Carpentry spinoff focused on software skills for librarians.

Key Points

• Humans are good at categorizing information, computers not so much.

• Often, data on a web site is not properly structured, making its extraction difficult.

• Web scraping is the process of automating the extraction of data from web sites.

• Web scraping is, in general, legal and won’t get you into trouble.

• There are a few things to be careful about, notably don’t overwhelm a web server and don’t steal content.

• Be nice. In doubt, ask.