Errata

The source "fdist1.plot(50,cumulative=True) gives a Y axis in counts and the label is "Cumulative Counts" rather than a Y axis in percentage and a label of "Cumulative Percentage". This using Python 2.6.2 and the nltk and plotting packages downloaded on 9/13/09.

Note from the Author or Editor:
This has been addressed in second printing.

Argument tuple to nltk.ConditionalFreqDist should be

(target,fileid[:4])

not

(target,file[:4])

.

Note from the Author or Editor:
Please see:
http://code.google.com/p/nltk/issues/detail?id=417

Already fixed in online version

More contrast (supplied image was color)

In the section on processing rss feeds, a line reads:

>>> nltk.word_tokenize(nltk.html_clean(content))

However, html_clean should be clean_html

In the section on processing rss feeds, a line reads:

>>> nltk.word_tokenize(nltk.html_clean(content))

However, html_clean should be clean_html

"for line in b: print b" should have been "for line in b: print line".

Note from the Author or Editor:
Please see:
http://code.google.com/p/nltk/issues/detail?id=418

Fixed in online version. Will be fixed in next printing.

"The final step is to search for the pattern of zeros and ones that maximizes this objective function, shown in Example 3.10. "
Shouldn't it be "The final step is to search for the pattern of zeros and ones that MINIMIZES this objective function..."? At least that's what the anneal function does...

Note from the Author or Editor:
Addressed in second printing.

[len(w) for w in nltk.corpus.brown.sents(categories='news'))]

should be

[len(w) for w in nltk.corpus.brown.sents(categories='news')]

Note from the Author or Editor:
Agreed.

There is no need to use nltk.defaultdict in this example.
The following code works fine.

#trie = nltk.defaultdict(dict)
trie = {}
insert(trie, 'chat', 'cat')
insert(trie, 'chien', 'dog')
insert(trie, 'chair', 'flesh')
insert(trie, 'chic', 'stylish')

#trie = dict(trie) # for nicer printing
trie['c']['h']['a']['t']['value']
pprint.pprint(trie)

Note from the Author or Editor:
I agree. The line:

trie = nltk.defaultdict(dict)

should be changed to:

trie = {}

>>> statement = "random.randint(0, %d) in vocab" % vocab_size * 2

This line should be corrected as follows:

>>> statement = "random.randint(0, %d) in vocab" % (vocab_size * 2)

Note from the Author or Editor:
Confirmed.

The nltk.corpus.wordnet object does not have path_distance().
It must be path_similarity().

Note from the Author or Editor:
The exercise should be changed to specify shortest_path_distance() instead of path_distance().

The line
>>> print nltk.ConfusionMatrix(gold, test)
causes an error.
It should be
>>> print nltk.ConfusionMatrix(gold_tags, test_tags)

Note from the Author or Editor:
Agreed. The line:

print nltk.ConfusionMatrix(gold, test)

should be:

print nltk.ConfusionMatrix(gold_tags, test_tags)

In example 7.4, the last statement in the parse method calls the method nltk.chunk.conlltags2tree. In NLTK 2.0.4 nltk.chunk has no such method. The call should be to nltk.chunk.util.conlltags2tree.

Note from the Author or Editor:
This error has been fixed by changing the imports at the level of the nltk.chunk package.

There is a method call to nltk.sem.show_raw_rtuple, which doesn't exist in NLTK 2.0.4. The call should be nltk.sem.relextract.show_raw_rtuple.

Note from the Author or Editor:
The interface to this module has been updated (see https://raw.github.com/nltk/nltk/master/ChangeLog), and the source files of the book have been revised accordingly.

Fig 9-1 is too big in the latest pdf. Also, the feature labels shouldn't be bold.

Remove box from code block

Within the code following the the sentence "This allows us to parse a query into SQL:", there is a line that assigns a value to the 'answer' variable:

answer = trees[0].node['sem']

When I try to enter this line, I get a stack trace followed by this error message:

KeyError: 'sem'

When I enter

answer = trees[0].node['SEM']

I get the prompt back, without any stack trace or error message.

Note from the Author or Editor:

It should say:

answer = trees[0].node['SEM']

Addressed in 2nd printing.

node['sem'] s/b node['SEM']
NB This is http://www.oreillynet.com/cs/nl/edit/errata/40392

Smaller scale

In the following two lines, the string "?subj" should be replaced with
"?np" (two substitutions):

(30) S[SEM=<?vp(?np)>] -> NP[SEM=?subj] VP[SEM=?vp]

(30) tells us that given some sem value ?subj for the subject NP and
some sem value ?vp for the VP

The string "?subj" should be replaced with "?np" (three substitutions):

S[SEM=[core=<?vp(?subj)>, store=(?b1+?b2)]] ->
NP[SEM=[core=?subj, store=?b1]] VP[SEM=[core=?vp, store=?b2]]

The core value at the S node is the result of applying the VP's core value, namely \x.smile(x), to the subject NP's value. The latter will not be @x, but rather an instantiation of @x, say z3. After &#946;-reduction, <?vp(?subj)> will be unified with <smile(z3)>.

Please replace all seven occurrences of
"nltk.ApplicationExpression" with "nltk.sem.ApplicationExpression".