How to get plain text out of wikipedia

问题:

I've been searching for about 2 months now to find a script that gets the Wikipedia description section only. (It's for a bot i'm building, not for IRC.) That is, when I say

/wiki bla bla bla

it will go to the Wikipedia page for bla bla bla, get the following, and return it to the chatroom:

"Bla Bla Bla" is the name of a song made by Gigi D'Agostino. He described this song as "a piece I wrote thinking of all the people who talk and talk without saying anything". The prominent but nonsensical vocal samples are taken from UK band Stretch's song "Why Did You Do It"

Here is the closest I've found, but it only gets the URL:

import json
import urllib.request, urllib.parse

def google(searchfor):
  query = urllib.parse.urlencode({'q': searchfor})
  url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&%s' % query

  search_response = urllib.request.urlopen(url)
  search_results = search_response.read().decode("utf8")
  results = json.loads(search_results)
  data = results['responseData']
  hits = data['results']

  if len(hits) > 0:
    return hits[0]['url']
  else:
    return "No results found."

(Python 3.1)

回答1:

Use the MediaWiki API, which runs on Wikipedia. You will have to do some parsing of the data yourself.

For instance:

http://en.wikipedia.org/w/api.php?action=query&prop=revisions&rvprop=content&format=json&&titles=Bla%20Bla%20Bla

means

fetch (action=query) the content (rvprop=content) of the most recent revision of Main Page (title=Main%20Page) in JSON format (format=json).

You will probably want to search for the query and use the first result, to handle spelling errors and the like.

回答2:

Here are a few different possible approaches; use whichever works for you. All my code examples below use requests for HTTP requests to the API; you can install requests with pip install requests if you have Pip. They also all use the Mediawiki API, and two use the query endpoint; follow those links if you want documentation.

1. Get a plain text representation of either the entire page or the page "extract" straight from the API with the extracts prop

Note that this approach only works on MediaWiki sites with the TextExtracts extension. This notably includes Wikipedia, but not some smaller Mediawiki sites like, say, http://www.wikia.com/

You want to hit a URL like

https://en.wikipedia.org/w/api.php?action=query&format=json&titles=Bla_Bla_Bla&prop=extracts&exintro&explaintext

Breaking that down, we've got the following parameters in there (documented at https://www.mediawiki.org/wiki/Extension:TextExtracts#query+extracts):

• action=query, format=json, and title=Bla_Bla_Bla are all standard MediaWiki API parameters
• prop=extracts makes us use the TextExtracts extension
• exintro limits the response to content before the first section heading
• explaintext makes the extract in the response be plain text instead of HTML

Then parse the JSON response and extract the extract:

>>> import requests
>>> response = requests.get(
...     'https://en.wikipedia.org/w/api.php',
...     params={
...         'action': 'query',
...         'format': 'json',
...         'titles': 'Bla Bla Bla',
...         'prop': 'extracts',
...         'exintro': True,
...         'explaintext': True,
...     }
... ).json()
>>> page = next(iter(response['query']['pages'].values()))
>>> print(page['extract'])
"Bla Bla Bla" is the title of a song written and recorded by Italian DJ Gigi D'Agostino. It was released in May 1999 as the third single from the album, L'Amour Toujours. It reached number 3 in Austria and number 15 in France. This song can also be heard in an added remixed mashup with L'Amour Toujours (I'll Fly With You) in its US radio version.

2. Get the full HTML of the page using the parse endpoint, parse it, and extract the first paragraph

MediaWiki has a parse endpoint that you can hit with a URL like https://en.wikipedia.org/w/api.php?action=parse&page=Bla_Bla_Bla to get the HTML of a page. You can then parse it with an HTML parser like lxml (install it first with pip install lxml) to extract the first paragraph.

For example:

>>> import requests
>>> from lxml import html
>>> response = requests.get(
...     'https://en.wikipedia.org/w/api.php',
...     params={
...         'action': 'parse',
...         'page': 'Bla Bla Bla',
...         'format': 'json',
...     }
... ).json()
>>> raw_html = response['parse']['text']['*']
>>> document = html.document_fromstring(raw_html)
>>> first_p = document.xpath('//p')[0]
>>> intro_text = first_p.text_content()
>>> print(intro_text)
"Bla Bla Bla" is the title of a song written and recorded by Italian DJ Gigi D'Agostino. It was released in May 1999 as the third single from the album, L'Amour Toujours. It reached number 3 in Austria and number 15 in France. This song can also be heard in an added remixed mashup with L'Amour Toujours (I'll Fly With You) in its US radio version.

3. Parse wikitext yourself

You can use the query API to get the page's wikitext, parse it using mwparserfromhell (install it first using pip install mwparserfromhell), then reduce it down to human-readable text using strip_code. strip_code doesn't work perfectly at the time of writing (as shown clearly in the example below) but will hopefully improve.

>>> import requests
>>> import mwparserfromhell
>>> response = requests.get(
...     'https://en.wikipedia.org/w/api.php',
...     params={
...         'action': 'query',
...         'format': 'json',
...         'titles': 'Bla Bla Bla',
...         'prop': 'revisions',
...         'rvprop': 'content',
...     }
... ).json()
>>> page = next(iter(response['query']['pages'].values()))
>>> wikicode = page['revisions'][0]['*']
>>> parsed_wikicode = mwparserfromhell.parse(wikicode)
>>> print(parsed_wikicode.strip_code())
{{dablink|For Ke$ha's song, see Blah Blah Blah (song). For other uses, see Blah (disambiguation)}}

"Bla Bla Bla" is the title of a song written and recorded by Italian DJ Gigi D'Agostino. It was released in May 1999 as the third single from the album, L'Amour Toujours. It reached number 3 in Austria and number 15 in France. This song can also be heard in an added remixed mashup with L'Amour Toujours (I'll Fly With You) in its US radio version.

Background and writing
He described this song as "a piece I wrote thinking of all the people who talk and talk without saying anything". The prominent but nonsensical vocal samples are taken from UK band Stretch's song "Why Did You Do It"''.

Music video
The song also featured a popular music video in the style of La Linea. The music video shows a man with a floating head and no arms walking toward what appears to be a shark that multiplies itself and can change direction. This style was also used in "The Riddle", another song by Gigi D'Agostino, originally from British singer Nik Kershaw.

Chart performance
Chart (1999-00)PeakpositionIreland (IRMA)Search for Irish peaks23

References

External links


Category:1999 singles
Category:Gigi D'Agostino songs
Category:1999 songs
Category:ZYX Music singles
Category:Songs written by Gigi D'Agostino

回答3:

You can fetch just the first section using the API:

http://en.wikipedia.org/w/api.php?action=query&prop=revisions&rvsection=0&titles=Bla%20Bla%20Bla&rvprop=content

This will give you raw wikitext, you'll have to deal with templates and markup.

Or you can fetch the whole page rendered into HTML which has its own pros and cons as far as parsing:

http://en.wikipedia.org/w/api.php?action=parse&prop=text&page=Bla_Bla_Bla

I can't see an easy way to get parsed HTML of the first section in a single call but you can do it with two calls by passing the wikitext you receive from the first URL back with text= in place of the page= in the second URL.

UPDATE

Sorry I neglected the "plain text" part of your question. Get the part of the article you want as HTML. It's much easier to strip HTML than to strip wikitext!

回答4:

You can get wiki data in Text formats. If you need to access many title's informations, you can get all title's wiki data in a single call. Use pipe character ( | ) to separate each titles.

http://en.wikipedia.org/w/api.php?format=json&action=query&prop=extracts&exlimit=max&explaintext&exintro&titles=Yahoo|Google&redirects=

Here this api call return both Googles and Yahoos data.

explaintext => Return extracts as plain text instead of limited HTML.

exlimit = max (now its 20); Otherwise only one result will return.

exintro => Return only content before the first section. If you want full data, just remove this.

redirects= Resolve redirect issues.

回答5:

DBPedia is the perfect solution for this problem. Here: http://dbpedia.org/page/Metallica, look at the perfectly organised data using RDF. One can query for anything here at http://dbpedia.org/sparql using SPARQL, the query language for the RDF. There's always a way to find the pageID so as to get descriptive text but this should do for the most part.

There will be a learning curve for RDF and SPARQL for writing any useful code but this is the perfect solution.

For example, a query run for Metallica returns an HTML table with the abstract in several different languages:

<table class="sparql" border="1">
  <tr>
    <th>abstract</th>
  </tr>
  <tr>
    <td><pre>"Metallica is an American heavy metal band formed..."@en</pre></td>
  </tr>
  <tr>
    <td><pre>"Metallica es una banda de thrash metal estadounidense..."@es</pre></td>
... 

SPARQL QUERY :

PREFIX dbpedia-owl: <http://dbpedia.org/ontology/>
PREFIX dbpprop: <http://dbpedia.org/property/>
PREFIX dbres: <http://dbpedia.org/resource/>

SELECT ?abstract WHERE {
 dbres:Metallica dbpedia-owl:abstract ?abstract.
}

Change "Metallica" to any resource name (resource name as in wikipedia.org/resourcename) for queries pertaining to abstract.

回答6:

I think the better option is to use the extracts prop that provides you MediaWiki API. It returns you only some tags (b, i, h#, span, ul, li) and removes tables, infoboxes, references, etc.

http://en.wikipedia.org/w/api.php?action=query&prop=extracts&titles=Bla%20Bla%20Bla&format=xml gives you something very simple:

<api><query><pages><page pageid="4456737" ns="0" title="Bla Bla Bla"><extract xml:space="preserve">
<p>"<b>Bla Bla Bla</b>" is the title of a song written and recorded by Italian DJ Gigi D'Agostino. It was released in May 1999 as the third single from the album, <i>L'Amour Toujours</i>. It reached number 3 in Austria and number 15 in France. This song can also be heard in an added remixed mashup with <i>L'Amour Toujours (I'll Fly With You)</i> in its US radio version.</p> <p></p> <h2><span id="Background_and_writing">Background and writing</span></h2> <p>He described this song as "a piece I wrote thinking of all the people who talk and talk without saying anything". The prominent but nonsensical vocal samples are taken from UK band Stretch's song <i>"Why Did You Do It"</i>.</p> <h2><span id="Music_video">Music video</span></h2> <p>The song also featured a popular music video in the style of La Linea. The music video shows a man with a floating head and no arms walking toward what appears to be a shark that multiplies itself and can change direction. This style was also used in "The Riddle", another song by Gigi D'Agostino, originally from British singer Nik Kershaw.</p> <h2><span id="Chart_performance">Chart performance</span></h2> <h2><span id="References">References</span></h2> <h2><span id="External_links">External links</span></h2> <ul><li>Full lyrics of this song at MetroLyrics</li> </ul>
</extract></page></pages></query></api>

You can then run it through a regular expression, in JavaScript would be something like this (maybe you have to do some minor modifications:

/^.*<\s*extract[^>]*\s*>\s*((?:[^<]*|<\s*\/?\s*[^>hH][^>]*\s*>)*).*<\s*(?:h|H).*$/.exec(data)

Which gives you (only paragrphs, bold and italic):

"Bla Bla Bla" is the title of a song written and recorded by Italian DJ Gigi D'Agostino. It was released in May 1999 as the third single from the album, L'Amour Toujours. It reached number 3 in Austria and number 15 in France. This song can also be heard in an added remixed mashup with L'Amour Toujours (I'll Fly With You) in its US radio version.

回答7:

"...a script that gets the Wikipedia description section only..."

For your application you might what to look on the dumps, e.g.: http://dumps.wikimedia.org/enwiki/20120702/

The particular files you need are 'abstract' XML files, e.g., this small one (22.7MB):

http://dumps.wikimedia.org/enwiki/20120702/enwiki-20120702-abstract19.xml

The XML has a tag called 'abstract' which contain the first part of each article.

Otherwise wikipedia2text uses, e.g., w3m to download the page with templates expanded and formatted to text. From that you might be able to pick out the abstract via a regular expression.

回答8:

You can try WikiExtractor: http://medialab.di.unipi.it/wiki/Wikipedia_Extractor

It's for Python 2.7 and 3.3+.

回答9:

First check here.

There are a lot of invalid syntaxes in MediaWiki's text markup. (Mistakes made by users...) Only MediaWiki can parse this hellish text. But still there are some alternatives to try in the link above. Not perfect, but better than nothing!

回答10:

You can try the BeautifulSoup HTML parsing library for python,but you'll have to write a simple parser.

回答11:

There is also the opportunity to consume Wikipedia pages through a wrapper API like JSONpedia, it works both live (ask for the current JSON representation of a Wiki page) and storage based (query multiple pages previously ingested in Elasticsearch and MongoDB). The output JSON also include plain rendered page text.