WebSearcher

Tools for conducting and parsing web searches

This package provides tools for conducting algorithm audits of web search and includes a scraper built on requests with tools for geolocating, conducting, and saving searches. It also includes a modular parser built on BeautifulSoup for decomposing a SERP into list of components with categorical classifications and position-based specifications.

Version numbers correspond to the date of the data corpus that the parser was last optimized to perform on. For example, version 2022.12.18 was developed using data collected on December 18, 2022. Due to the rate at which HTML changes over time, the parsers built on them frequently need updating. This version scheme should help to select which version to use for historical datasets going forward. Earlier versions (0.1.2 to 0.2.15) were optimized for various periods from 2018 to 2020. The last version with this original numbering (0.2.15) was optimized on mid-to-late 2020 data, and is also available as version 2020.0.0.

WebSearcher

Getting Started

# Install pip version
pip install WebSearcher

# Install Github development version - less stable, more fun!
pip install git+https://github.com/gitronald/WebSearcher@dev

Usage

Prepare a search

import WebSearcher as ws

# Initialize crawler with defaults (headers, logs, ssh tunnels)
se = ws.SearchEngine()
vars(se)

{'url': 'https://www.google.com/search',
 'params': {},
 'headers': {'Host': 'www.google.com',
  'Referer': 'https://www.google.com/',
  'Accept': '*/*',
  'Accept-Language': 'en-US,en;q=0.5',
  'Accept-Encoding': 'gzip,deflate,br',
  'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:58.0) Gecko/20100101 Firefox/58.0'},
 'ssh_tunnel': None,
 'sesh': <requests.sessions.Session at 0x7f7bad8efba8>,
 'log': <Logger WebSearcher.searchers (DEBUG)>,
 'html': None,
 'results': [],
 'results_html': []}

Conduct a search

# Conduct Search
se.search('immigration')

2019-08-14 01:25:38,267 | 2688 | INFO | WebSearcher.searchers | 200 | Searching immigration

# Parse Results
se.parse_results()

2019-08-14 01:25:42,208 | 2688 | INFO | WebSearcher.parsers | Parsing SERP 4d4fe27fe6b6466041e326622719b03ccc6542427c577c69740ae7fc

se.results[0]
{'cite': 'The New York Times',
 'cmpt_rank': 0,
 'details': {'img_url': None, 'live_stamp': False, 'orient': 'h'},
 'lang': 'en',
 'qry': 'immigration',
 'serp_id': '4d4fe27fe6b6466041e326622719b03ccc6542427c577c69740ae7fc',
 'serp_rank': 0,
 'sub_rank': 0,
 'timestamp': '1 day ago',
 'title': 'Trump Policy Favors Wealthier Immigrants for Green Cards',
 'type': 'top_stories',
 'url': 'https://www.nytimes.com/2019/08/12/us/politics/trump-immigration-policy.html'}

Save a Search

Recommended: Append html and meta data as lines to a json file. Useful for larger or ongoing crawls.

se.save_serp(append_to='serps.json')

Alternative: Save individual html files in a directory, named by a provided or (default) generated serp_id. Useful for smaller qualitative explorations where you want to quickly look at what is showing up. No meta data is saved, but timestamps could be recovered from the files themselves.

se.save_serp(save_dir='./serps')

Localization

Conduct web searches from a location of choice.

Obtain location names

To conduct localized searches, you only need one additional data point: The "Canonical Name" of each location.
These are available online, and can be downloaded using a built in function that checks for the most recent version.

# Set save location  
data_dir = './location_data'

# Download latest data; checks for local version before downloading
ws.download_locations(data_dir)

f  = os.listdir(data_dir)[-1]  # Last file
fp = os.path.join(data_dir, f) # File path
locs = pd.read_csv(fp)         # Read
locs.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 102029 entries, 0 to 102028
Data columns (total 7 columns):
Criteria ID       102029 non-null int64
Name              102029 non-null object
Canonical Name    102029 non-null object
Parent ID         101788 non-null float64
Country Code      102013 non-null object
Target Type       102029 non-null object
Status            102029 non-null object
dtypes: float64(1), int64(1), object(5)
memory usage: 5.4+ MB

# Take a look at the first row
locs.iloc[0]

Criteria ID                       1000002
Name                                Kabul
Canonical Name    Kabul,Kabul,Afghanistan
Parent ID                     9.07539e+06
Country Code                           AF
Target Type                          City
Status                             Active
Name: 0, dtype: object

Looking for Canonical Names. In order to search from a given location, you must find the corresponding canonical name.

# Filter for names containing "Boston" and "Massachusetts"
regex = r'(?=.*Boston)(?=.*Massachusetts)' 
str_mask = locs['Canonical Name'].str.contains(regex)
locs[str_mask]

15849                                Boston,Massachusetts,United States
15908                           East Boston,Massachusetts,United States
66033    Boston Logan International Airport,Massachusetts,United States
84817                        Boston College,Massachusetts,United States
85985                          South Boston,Massachusetts,United States
Name: Canonical Name, dtype: object

Conduct a localized search

After picking one, say 'Boston,Massachusetts,United States', you just add this to your se.search() call:

# Conduct Search
qry = 'pizza'
loc = 'Boston,Massachusetts,United States'
se.search(qry, location=loc)

# Parse Results
se.parse_results()

# Shape as dataframe
results = pd.DataFrame(se.results)

# Show details of local results returned 
results[results.type=='local_results']['details'].tolist()

[{
    'rating': 4.0,
    'n_reviews': 152,
    'sub_type': 'Pizza',
    'contact': '226 N Market St'
},
{
    'rating': 4.6,
    'n_reviews': 752,
    'sub_type': 'Pizza',
    'contact': '69 Salem St'
},
{
    'sub_type': 'Pizza', 
    'contact': 'McCormack Building, 1 Ashburton Pl'
}]

Contributing

Happy to have help! If you see a component that we aren't covering yet, please add it using the process below. If you have other improvements, feel free to add them any way you can.

Coming next:
- Functions for using multiprocessing to parse SERPs.

Repair or Enhance a Parser

Examine parser names in /component_parsers/__init__.py
Find parser file as /component_parsers/{cmpt_name}.py.

Add a Parser

Add classifier to component_classifier.py, as 'cmpt_name'
Add parser file in /component_parsers as cmpt_name.py, with function parse_cmpt_name.
Add import for parse_cmpt_name in /component_parsers/__init__.py

Recent Changes

0.2.15 - Fix people-also-ask and hotel false positives, add flag for left-hand side bar

0.2.14 - Add shopping ads carousel and three knowledge subtypes (flights, hotels, events)

0.2.13 - Small fixes for knowledge subtypes, general subtypes, and ads

0.2.12 - Try to brotli decompress by default

0.2.11 - Fixed local result parser and no return in general extra details

0.2.10 - a) Add right-hand-side knowledge panel and top image carousel, b) Add knowledge and general component subtypes, c) Updates to component classifier, footer, ad, and people_also_ask components

0.2.9 - Various fixes for SERPs with a left-hand side bar, which are becoming more common and change other parts of the SERP layout.

0.2.8 - Small fixes due to HTML changes, such as missing titles and URLs in general components

0.2.7 - Added fix for parsing twitter cards, removed pandas dependencies and several unused functions, moving towards greater package simplicity.

0.2.6 - Updated ad parser for latest format, still handles older ad format.

0.2.5 - Google Search, like most online platforms, undergoes changes over time. These changes often affect not just their outward appearance, but the underlying code that parsers depend on. This makes parsing a goal with a moving target. Sometime around February 2020, Google changed a few elements of their HTML structure which broke this parser. I created this patch for these changes, but have not tested its backwards compatibility (e.g. on SERPs collected prior to 2/2020). More generally, there's no guarantee on future compatibility. In fact, there is almost certainly the opposite: more changes will inevitably occur. If you have older data that you need to parse and the current parser doesn't work, you can try using 0.2.1, or send a pull request if you find a way to make both work!

Similar Packages

Many other methods for scraping web search in python exist, but many of these projects have been abandoned, and I was not able to find one that provided details on the components (e.g. "Answer Boxes" and "Top Stories") and positioning configurations (e.g. a horizontally or vertically oriented carousel) that compose a modern SERP.

Some of the other projects are still ongoing and very interesting in their own ways. Feel free to add to the list through a pull request if you are aware of others:

License

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

sorokinvld / websearcher Goto Github PK

websearcher's Introduction

WebSearcher

Tools for conducting and parsing web searches

Table of Contents