keighrim / banananer Goto Github PK

indicate the zone in the article
possible values: TXT, HL, DATELINE, DD

Here are some ideas for features to implement:

Dictionaries of named entities (pulled from net)
Clusters
POS tags
Word > average length of all words
Word > 10 characters (or some other length)
BIO chunks included in dictionary of names or not

• Unigram, Bigram, Trigram, Quadrigram included in dictionary of names

Word capitalized or not
Number of capitalized letters in word
Number of vowels in word
Number of consonants in word
Specific letters?
Length of sentence
First half/second half of sentence
Word is 'banana' or not

1/26/2015

• Base code for feature extraction - Yalin
• Base code for an interface between CRF++ - Keigh
• Get Brown clustering running - Todd

We don't have any bug, since we don't have any code

combination of CASE and ZONE

http://web.archive.org/web/20140208005616/http://metaoptimize.com/projects/wordreprs/

original site is not configured

if token is the first word in a sentence
possible vales: bool

zone and case of the previous token

zone and case of the next token

Need a class that has at least:

1. read() - a method to load up train.gold file into a list of sentences (or any other data structure)
2. any necessary method that return tokens, sentences, <token, postag> pairs at token level, <token, postag> pairs at sentence level
3. feature_functions = [] - a list of names of functions, doing one job for each. Since we already have POS tags, something like pos_tag() can be the first element of this list
4. extract() - method to traverse all functions in feature_functions to create feature vector

feature vector data structure

Basically it should be a table-like structure, where each column represents each feat_func in feat_funcs, and each row is for each token

Any additions or suggestions are welcomed.

possible values: InitCaps, MixedCaps, AllCaps

I get a question for the use of url. In the url, it did not give us the names and it is just the website. I do not know how to use it. I tried

    file = urllib2.urlopen('http://www.timeanddate.com/')
    data = file.read()
    file.close()
    tree=ET.parse(file)
    root=tree.getroot()

I did not get the correct root. I am confused.