2

For a project I would like to look up the top bigrams containing the word "like". Is there such a service available online somewhere?

  • It seems you want bigrams of words and not of letters or any other level. Is that correct? – hippietrail May 24 '13 at 11:47
  • 1
    Yes, I'm mostly looking for word tokens. – Jimmy C May 24 '13 at 12:55
  • Do you have a preference as to what type of language you wish to query? News articles, tweets, etc. – acattle May 25 '13 at 3:59
  • 1
    @JimmyCallin: Do you want this lemmatized? Meaning do you want all forms of like treated the same: liked, likes, liking. Do you care about part-of-speech? Like is mainly either a verb or a preposition (but it's actually more complicated than that) - do you want those mixed up or separated? "He likes food like hamburgers." The first like is an inflected verb, the second is a preposition. Each will have different n-gram characteristics but separating them is not trivial. – hippietrail May 25 '13 at 5:26
1

(This is mostly a step-by-step explanation of how to make use of the sources acattle pointed out.)

The web interface of the Corpus of Contemporary American English offers a collocate search function. Just follow the link and you should see the web interface. Now choose 'Compare' under 'Display', this should make a 'Collocate' box appear immediately under the search box. Then choose 'List' under 'Display', and type in 'like' in the search box. Now press search and you should get the major collocates of 'like' and their frequency.

Mark Davies also offers other corpora which you can access with the same interface. If that does not suit your needs you could get another corpus (for example from the sources acattle pointed out) and extract collocates/collocations with a concordancer (a program for accessing/searching corpora). I would recommend the free program Antconc.

1

You can try the Google Ngram Viewer. They additionally provide the raw data for analysis.

0

I found this website which might meet your needs but I honestly can't figure out exactly how you'd use it.

If that doesn't work, it's easy enough to write your own program which reads a corpus, counts the bigrams, and outputs the data you're looking for. Even if you don't have any programming background, this is actually quite easy. I have given this task as an assignment to classes where the majority of students had little-to-no programming experience and that didn't have much trouble. Try a quick Google search for tutorials.

The problem is choosing a corpus. Generally corpora tend to be specialized by the type of language included (such as only news articles, only tweets, only novels, etc.). You will need to take a look at what's available and decide which corpus (or combination of corpora) are right for you.

This website offers multiple corpora for free download as well as TREC offers multiple corpora across various types of language (some free, some requiring approval, and some requiring purchase).

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.