I have a huge amount of documents (mainly pdfs and doc's) I want to classify, so I can search over them according to certain tags. These tags could either be of my own (I put the tags to the document) or extracted from the text.
I've just seen a post related to this (Classify data using Apache Mahout), but perhaps there is something even more simple.
Mahout might be overkill for your problem - but you can get a fairly quick, easy solution by using OpenNLP.
http://opennlp.sourceforge.net/api/index.html
Specifically, look at the opennlp.tools.doccat package. Essentially, you have to go through and manually tag a small(ish) set of the items for each category you desire. If they are really distinct, you can get away with a small sample size.
You can use the DocumentCategorizerME.train() static function to train a collection of documents, where each requires a category tag and the text block to train on. Then, you can initialize the DocumentCategorizerME with the trained model and begin classifying all the rest of your documents.
Once you do this, you can (I think) write the model to a file so you don't have to ever do that again.
This post on extracting keywords and classifying webpages is related and may be helpful. In your example it sounds like you can use tags in lieu of the keyword extraction piece (although you may want to use both in combination). Weka is easy to use, I would definitely recommend giving it a look.
Related
I need to create a universal web scraper to parse articles on the different websites. Of course, I know about XPath, but I want to try to make it universal for any website despite the HTML markup of a page.
I need to determine whether there is an article on the page and if it is - parse a text of title, body and tags (if exists).
Frankly speaking, my knowledge in DS is not very huge, but I assume this task (determine whether it is article, and parsing only needed parts) is possible to solve.
What tools should I use? Any help?
Actually, for the second task, I need to implement something similar that google chrome mobile does. When page is not optimised for mobile, then propose to show the page in adaptive mode (just title, and main content).
If you are using Python, some libraries to look at are:
scrapy, which scrapes data and can extract some of the results) and,
BeautifulSoup, which is more geared towards the extraction part itself.
It is possible to request a version of a website (e.g. for Chrome, Safari, Mobile, old-school systems) by creating a custom header for your scraper.
HAve a look at the relevant documentation, and you can get an idea of how to use headers in scrapy here.
I do not know of any more specialised tools. Your tasks are more analytical and are typically not performed with the use of models for estimating e.g. what content is where on a webpage. This might be an intersting research direction though; to see if you can create a model that generalises across many websites to extract the desired content.
That leads me on to my last point, which is to say that creating a single scraper that works for any website *containing your artile type) is not usually possible. People create websites differently, however they see fit, which means they also change them. This usually leads to a good scraper requiring constant updates as time (and developers) moves on.
EDIT:
Then if you have lots of labelled examples, it might be possible to train a model. The challenge might be the look-back range of the model. For example, a typical LSTM model is given a parameter that tells it how far to look back into the past. It is stored within its memory internally. In your case, you might be looking for a start and end HTML tag of an article, to then extract just that part. These tahs could be thousands of words apart. Something a standard LSTM might not be fit to retain and use.
If you could pose your problem a little differently, then there are other approaches that might be plausible. E.g., you could make it a "question-answer" problem, by saying: I have this HTML, where is the article content? If that sounds ok for your use-case, have a look here for some model based approaches.
I'm new to OSM querying, but would like to query vector data for a large area. Thus I need to limit the results I would like to get by tagging the request.
http://www.informationfreeway.org/api/0.6/way[tag=value][bbox=x,y,z,j]
I'd like to filter for specific tag/values when querying for a way. Though I don't know which tags/values exist. Is there a list listing the most common of them?
You are approaching your problem from the wrong direction. The number of different tags is almost unlimited. According to taginfo there are currently 75 380 856 different tags. I'm pretty sure you are not interested in most of them. Likewise you are probably not even interested in many of the most common tags.
What data do you want to query?
The OSM wiki should be your starting point for generating a list of tags you are interested in. For a generic overview take a look at the map features. Are you interested in streets? Then visit at the highway key. Routing? Then take a look at the routing wiki page.
Always remember that these lists aren't complete. People can use any tag they like (but should use well-established tags whenever possible of course).
Also consider using Overpass API instead of XAPI. Overpass API is much more powerful.
I'm building a site that allows users to make sense of a debate by graphically representing arguments for and against a particular issue. (Wrangl)
I'd like to categorise these debates so they are more easily found and connected. I don't want to irritate the person creating the debate by asking them to add tags and categories before they see any benefit, so I'm looking at a way of automatically extracting keywords.
What's a good approach for taking the debate's title and description (and possibly the content of the arguments themselves once there are some) to pull out, say, ten strong keywords that could be used as metadata to connect similar debates together, or even as the content of the "meta" keywords tag in the head of the HTML page where the debate is viewable. Eg. Datamapper vs ActiveRecord
The site is coded in Ruby with Sinatra, using DataMapper for data storage. I'm ideally looking for something which will work on Heroku (I don't have a way of writing files to disk dynamically), and I'd consider a web service, an API or ideally a Ruby gem.
Maybe you can use TextAnalyzer.
I understand that you're wanting to find an easy way of achieving this, I've recently dived into the world of NLP (Natural Language Processing) and Text-mining and its a daunting process of which most went far above my head.
Although i managed to code some functionality that resembles what you're looking for, though I did it in PHP. What i would suggest, that if you want it tailored to your project (Wrangl) then do it yourself.
Using the Porter stemming algorithm which I'm sure there will be Ruby code for.
Ruby Porter stemmer
You can try the salsaAPI to automatically extract keywords and categorize the debates!
I'm looking into Thinking Sphinx for it's potential to solve an indexing problem. It looks like it has a very specific API for telling it what fields to index on a model. I don't like having this layer of abstraction in my way without being able to sidestep it. The thing is I don't trust Sphinx to be able to interpret my model properly as this model could have any conceivable property. Basically, I want to encode JSON in a RDBMS. In a way, I'm looking to make an RDBMS behave like MongoDB (RDBMSes have features I don't want to do without). If TS or some other index could be made to understand my models this could work. Is it possible to manually provide key/value pairs to TS?
"person.name.first" => "John", "person.name.last" => "Doe", "person.age" => 32,
"person.address" => "123 Main St.", "person.kids" => ["Ed", "Harry"]
Is there another indexing tool that could be used from Ruby to index JSON?
(By the way, I have explored a wide variety of NoSQL databases. I am trying to address a very specific set of requirements.)
As Matchu has pointed out in the comments, Sphinx usually interacts directly with the database. This is why Thinking Sphinx is built like it is.
However, Sphinx (but not Thinking Sphinx) can also accept XML data formats - so if you want to go down that path, feel free. You're going to have to understand the underlying Sphinx structure much more deeply than you would if using a normal relational database/ActiveRecord and Thinking Sphinx approach. Riddle may be useful for building a solution, but you'll still need to understand Sphinx itself first.
Basically, when you're specifying what you want to index--that is, when you want to build your own index--you're using the Map part of Map/Reduce. CouchDB supports exactly this. The only problem I ran into with Couch is that I want to query other document objects as the basis of my Map/Reduce since those documents would contain metadata about how I want to build my indexes. This goes against the grain of Map/Reduce however as you have to map a document in isolation with no external data. If you need external data it would instead be denormalized into your documents.
I have to keep up with structured documents containing things such as requests for proposals, government program reports, threat models and all kinds of things like that. They are in techno-legalese as I would call them: highly structured, with section numbering and 3, 4 and 5 levels of nesting. All in English
I need a more efficient way to locate those paragraphs of nuggets that matter to me. So what I’d like is kind of a local document index/repository, that would allow me to have some standing queries and easily locate sections in documents that talk about my queries. Here’s an example:
I’d like to load in 10 large PDF files, each of say 100 pages. Each PDF contains English text, formatted very nicely into paragraphs and sections.
I’d like to specify that I am interested in “blogging platforms”, “weaknesses in Ruby”, “localization and internationalization”
Ideally then look at a list that showed the section of text, the name of the document, and other information that seemed to be related to and/or include the words and phrases I specified.
I am sure something like this exists. I would call it something like document indexing, document comprehension or structured searching.
Take a look at Lucene: http://lucene.apache.org/ and Solr http://lucene.apache.org/solr/ , which can do most of what you ask. They are not exaclty featherweight though!
There is also this excellent book:
http://www.amazon.com/Building-Search-Applications-Lucene-Lingpipe/dp/0615204252/
Opengrok is another lightweight solution on top of Lucene: http://opengrok.github.io/OpenGrok/
Alternatively, you could have a look at http://www.alfresco.com,
which is not lightweight solution but it is designed exactly for your purposes.