The Automatic Acquisition of Frequencies of Verb 
Subcategorization Frames from Tagged Corpora 
Akira Ushioda, David A. Evans, Ted Gibson, Alex Waibel 
Computational Linguistics Program 
Carnegie Mellon University 
Pittsburgh, PA 15P13-3890 
aushioda @icl. cmu. edu 
Abstract 
We describe a mechanism for automatically acquiring verb subcategorization frames 
and their frequencies in a large corpus. A tagged corpus is first partially parsed to 
identify noun phrases and then a finear grammar is used to estimate the appropri- 
ate subcategorization frame for each verb token in the corpus. In an experiment 
involving the identification of six fixed subcategorization frames, our current system 
showed more than 80% accuracy. In addition, a new statistical approach substan- 
tially improves the accuracy of the frequency estimation. 
1 Introduction 
When we construct a grammar, there is always a trade-off between the coverage of the 
grammar and the ambiguity of the grammar. If we hope to develop an efficient high- 
coverage parser for unrestricted texts, we must have some means of dealing with the 
combinatorial explosion of syntactic ambiguities. While a general probabilistic optimiza- 
tion technique such as the Inside-Outside algorithm (\[Baker, 1979\], \[Lauri and Young, 
1990\], \[Jelinek el ai., 1990\], \[Carroll and Charniak, 1992\]) can be used to reduce ambi- 
guity by providing estimates on the applicability of the context-free rules in a grammar 
(for example), the algorithm does not take advantage of lexical information, including 
such information as verb subcategorization frame preferences. Discovering or acquiring 
lexically-sensitive linguistic structures from large corpora may offer an essential comple- 
mentary approach. 
Verb subcategorization (verb-subcat) frames represent one of the most important ele- 
ments of grammatical/lexical knowledge for efficient and reliable parsing. At this stage in 
the computational-linguistic exploration of corpora, dictionaries are still probably more re- 
liable than automatic acquisition systems as a source of subcategorization (subcat) frames 
for verbs. The Oxford Advanced Learners Dictionary (OALD) \[Hornby, 1989\], for exam- 
ple, uses 32 verb patterns to describe a usage of each verb for each meaning of the verb. 
However, dictionaries do not provide quantitative information such as how often each 
verb is used with each of the possible subcat frames. Since dictionaries are repositories, 
primarily, of what is possible, not what is most likely, they tend to contain information 
about rare usage \[de Marken, 1992\]. But without information about the frequencies of the 
subcat frames we find in dictionaries, we face the prospect of having to treat each frame 
as equiprobable in parsing. This can lead to serious inefficiency. We also know that the 
frequency of subcat frames can vary by domain; frames that are very rare in one domain 
can be quite common in another. If we could automatically determine the frequencies 
of subcat frames for domains, we would be able to tailor parsing with domain-specific 
95 
heuristics. Indeed, it would be desirable to have a subcat dictionary for each possible 
domain. 
This paper describes a mechanism for automatically acquiring subcat frames and their 
frequencies based on a tagged corpus. The method utilizes a tagged corpus because (i) 
we don't have to deal with a lexical ambiguity (ii) tagged corpora in various domains 
are becoming readily available and (iii) simple and robust tagging techniques using such 
corpora recently have been developed (\[Church, 1988\], \[Brill, 1992\]). 
Brent reports a method for automatically acquiring subcat frames but without fre- 
quency measurements (\[Brent and Berwick, 1991\], \[Brent, 1991\]). His approach is to 
count occurrences of those unambiguous verb phrases that contain no noun phrases other 
than pronouns or proper nouns. By thus restricting the "features" that trigger identifi- 
cation of a verb phrase, he avoids possible errors due to syntactic ambiguity. Although 
the rate of false positives is very low in his system, his syntactic features are so selective 
that most verb tokens fail to satisfy them. (For example, verbs that occurred fewer than 
20 times in the corpus tend to have no co-occurrences with the features.) Therefore his 
approach is not useful in determining verb-subcat frame frequencies. 
To measure frequencies, we need, ideally, to identify a subcat frame for each verb token 
in the corpus. This, in turn, requires a full parse of the corpus. Since manually parsed 
corpora are rare and typically small, and since automatically parsed corpora contain 
many errors (given current parsing technologies), an alternative source of useful linguistic 
structure is needed. We have elected to use partially parsed sentences automatically 
derived from a lexically-tagged corpus. The partial parse contains information about 
minimal noun phrases (without PP attachment or clausal complements). While such 
derived information about linguistic structure is less accurate and complete than that 
available in certified, hand-parsed corpora, the approach promises to generalize and to 
yield large sample sizes. In particular, we can use partially parsed corpora to measure 
verb-subcat frame frequencies. 
2 Method 
The procedure to find verh-subcat frequencies, automatically, is as follows. 
(1) Make a list of verbs out of the tagged corpus. 
(2) For each verb on the list (the "target verb"), 
(2.1) Tokenize each sentence containing the target verb in the following way: 
All the noun phrases except pronouns are tokenized as "n" by a noun phrase 
parser and all the rest of the words are also tokenized following the schmema 
in Table 1. For example, the sentence "The corresponding mental-state verbs 
do not follow \[target verb\] these rules in a straightforward way" is transformed 
to a sequence of tokens "bnvaknpne'. 
(2.2) Apply a set of subcat extraction rules to the tokenized sentences. These rules 
are written as regular expressions and they are obtained through the examina- 
tion of occurrences of a small sample of verbs in a training text. 
Note that in the actual implementation of the procedure, all of the redundant oper- 
ations are eliminated. Our NP parser also uses a finite-state grammar. It is designed 
96 
b: sentence initial maker 
k: target verb 
i: pronoun 
n: noun phrase 
v: finite verb 
u: participial verb 
d: base form verb 
p: preposition 
e: sentence final maker 
t: "to" 
m: modal 
w: relative pronoun 
a: adverb 
x: punctuation 
c: complementizer "that" 
s: the rest 
Table 1: List of Symbols/Categories 
especially to support identification of verb-subcat frames. One of its special features is 
that it detects time-adjuncts such as "yesterday", "two months ago", or "the following 
day", and eliminates them in the tokenization process. For example, the sentence "He told 
the reporters the following day that..." is tokenized to "bivnc..." instead of "bivnnc...". 
3 Experiment on Wall Street Journal Corpus 
We used the above method in experiments involving a tagged corpus of Wall Street Journal 
(WSJ) articles, provided by the Penn Treebank project. Our experiment was limited in 
two senses. First, we treated all prepositional phrases as adjuncts. (It is generally difficult 
to distinguish complement and adjunct PPs.) Second, we measured the frequencies of 
only six fixed subcat frames for verbs in non-participle form. (This does not represent 
an essential shortcoming in the method; we only need to have additional subcat frame 
extraction rules to accommodate participles.) 
We extracted two sets of tagged sentences from the WSJ corpus, each representing 3- 
MBytes and approximately 300,000 words of text. One set was used as a training corpus, 
the other as a test corpus. Table 2 gives the list of verb-subcat frame extraction rules 
obtained (via examination) for four verbs "expect", "reflect", "tell", and "give", as they 
occurred in the training corpus. Sample sentences that can be captured by each set of 
rules are attached to the list. Table 3 shows the result of the hand comparison of the 
automat!cally identified verb-subcat frames for "give" and "expect" in the test corpus. 
The tabular columns give actual frequencies for each verb-subcat frame based on man- 
ual review and the tabular rowsgive the frequencies as determined automatically by the 
system. The count of each cell (\[i, j\]) gives the number of occurrences of the verb that 
are assigned the i-th subcat frame by the system and assigned the j-th frame by manual 
review. The frame/column labeled "REST" represents all other subcat frames, encom- 
passing such subcat frames as those involving wh-clauses, verb-particle combinations (such 
as "give up"), and no complements. 
Despite the simplicity of the rules, the frequencies for subcat frames determined under 
automatic processing are very close to the real distributions. Most of the errors are 
attributable to errors in the noun phrase parser. For example, 10 out of the 13 errors 
in the \[NP,NP+NP\] cell under "give" are due to noun phrase parsing errors such as the 
misidentification of a N-N sequence (e.g., *"give \[NP government officials rights\] against 
the press" vs. "give \[NP government officials\] \[NP rights\] against the press"). 
97 
Notes: 
NP: noun phrase 
Frame 
1. NP+NP 
2. NP+CL 
3. NPTINF 
4. CL 
5. NP 
6. INF 
Rule 
k(iln)n 
k(iln(pn)*)c 
k(iln)(iln)a*(mlv) 
k(iln(pn)*)ta*d 
k¢ 
k(iln)a*(mlv) 
k(iln)/\['mvd\] 
#pw(iln(pn)*)a*m?a*k/\['t\] 
kta*d 
CL: that-clause with and without the complementizer "that" 
INF: "to" + infinitive 
x* matches a sequence of any number of x's including zero x 
x? is either x or empty 
(xly) matches either x or y 
\['xyz\] matches any token except x, y, and z 
Ix(sequence) matches (sequence) that is not directly preceded by x 
x/y matches x if x is immediately followed by y 
Sample Sentences: 
Frame 1. 
Frame 2. 
Frame 3. 
~rame 4. 
Frame 5. 
Frame 6. 
"...gives current management enough time to work on..." 
"...tel_.._l the people in the hall that..." ; "...tol__.d him the man would..." 
"...expected the impact from the restructuring to make..." 
"...thlnk that..." ; "...thought the company eventually responded..." 
"...sa__.E the man..." ; "...which the president of the company wanted..." 
but not 
"...sa__~ him swim..."; "...(hotel) in which he stayed..."; "...(gift) which he expected to get..." 
"...expects to gain..." 
Table 2: Set of Subcategorization Frame Extraction Rules 
98 
NP-t-NP 
NP+CL Output 
NP+INF 
of NP 
System CL 
INF 
REST 
Total 
"Give" 
Real Occurrences 
NP+NP NP+CL NP+INF NP CL INF REST 
52 0 0 0 0 0 0 
l 0 0 0 0 0 0 
2 0 0 0 0 0 0 
13 0 0 27 0 0 0 
0 0 0 0 0 0 0 
0 0 0 0 0 0 0 
1 0 0 4 0 0 9 
69 0 0 31 0 0 9 
Total 
52 
1 
2 
40 
0 
0 
14 
109 
NP+NP 
NP+CL 
Output NP+INF 
of NP 
System CL 
INF 
REST 
Total 
"Expect" 
Real Occurrences 
NP+NP NP+CL NP+INF NP CL INF REST 
0 0 0 0 0 0 0 
0 0 0 0 0 0 0 
0 0 55 1 0 0 0 
0 0 4 28 0 0 0 
0 0 0 0 8 0 0 
0 0 0 0 0 40 0 
0 0 1 6 0 0 7 
Total 
0 
0 
56 
32 
8 
40 
14 
0 0 60 35 8 40 7 150 
Table 3: Subcategorization Frame Frequencies 
acquire end like 
build expand need 
close fail produce 
comment file prove 
consider follow reach 
continue get receive 
design help reduce 
develop hold see 
elect let sign 
spend 
total 
try 
use 
want 
work 
Table 4: Verbs Tested 
99 
THIS PAGE INTENTIONALLY LEFT BLANK 
I00 
Number 
of 
Verbs 
10-- 
6-- i 
3-- 
<S 
I l IV I 
5-10 10-15 15-20 20-25 25-30 30-35 34--40 40-45 
Error Rate (070) 
Figure 1: Distribution of Errors 
To measure the total accuracy of the system, we randomly chose 33 verbs from the 
300 most frequent verbs in the test corpus (given in Table 4), automatically estimated 
the subcat frames for each occurrence of these verbs in the test corpus, and compared the 
results to manually determined subcat frames. 
The overall results are quite promising. The total number of occurrences of the 33 
verbs in the test corpus (excluding participle forms) is 2,242. Of these, 1,933 were assigned 
correct subcat frames by the system. (The 'correct'-assignment counts always appear in 
the diagonal cells in a comparison table such as in Table 3.) This indicates an overall 
accuracy for the method of 86%. 
If we exclude the subcat frame "REST" from our statistics, the total number of oc- 
currences of the 33 verbs in one of the six subcat frames is 1,565. Of these, 1,311 were 
assigned correct subcat frames by the system. This represents 83% accuracy. 
For 30 of the 33 verbs, both the first and the second (if any) most frequent subcat 
frames as determined by the system were correct. For all of the verbs except one ("need"), 
the most frequent frame was correct. 
Figure 1 is a histogram showing the number of verbs within each error-rate zone. 
In computing the error rate, we divide the total 'off-diagonal'-cell counts, excluding the 
counts in the "REST" column, by the total cell counts, again excluding the "REST" col- 
umn margin. Thus, the off-diagonal cell counts in the "REST" row, representing instances 
where one of the six actual subcat frames was misidentified as "REST", are counted as 
errors. This formula, in general, gives higher error rates than would result from simply 
dividing the off-diagonal cell counts by the total cell counts. 
Overall, the most frequent source of errors, again, was errors in noun phrase boundary 
detection. The second most frequent source was misidentification of infinitival 'purpose' 
clauses, as in "he used a crowbar to open the door". "To open the door" is a 'purpose' 
adjunct modifying either the verb phrase "used a crowbar" or the main clause "he used a 
crowbar". But such adjuncts are incorrectly judged to be complements of their main verbs 
i01 
by the subcat frame extraction rules in Table 2. In formulating the rules, we assumed that 
a 'purpose' adjunct appears effectively randomly and much less frequently than infinitival 
complements. This is true for our corpus in general; but some verbs, such as "use" and 
"need", appear relatively frequently with 'purpose' infinitivals. In addition to errors from 
parsing and 'purpose' infinitives, we observed several other, less frequent types of errors. 
These, too, pattern with specific verbs and do not occur randomly across verbs. 
4 Statistical Analysis 
For most of the verbs in the experiment, our method provides a good measure of subcat 
frame frequencies. However, some of the verbs seem to appear in syntactic structures that 
cannot be captured by our inventory of subcat frames. For example, "need" is frequently 
used in relative clauses without relative pronouns, as in "the last thing they need". Since 
this kind of relative clauses cannot be captured by the rules in Table 2, each occurrence 
of these relative clause causes an error in measurement. It is likely that there are many 
other classes of verbs with distinctive syntactic preferences. If we try to add rules for each 
such class, it will become increasingly difficult to write rules that affect only the target 
class and to eliminate undesirable rule interactions. 
In the following sections, we describe a statistical method which, based on a set of 
training samples, enables the system to learn patterns of errors and substantially increase 
the accuracy of estimated verb-suhcat frequencies. 
4.1 General Scheme 
The method described in Section 2 is wholly deterministic; it depends only on one set 
of subcat extraction rules which serve as filters. Instead of treating the system output 
for each verb token as an estimated subcat frame, we can think of the output as one 
feature associated with the occurrence of the verb. This single feature can be combined, 
statistically, with other features in the corpus to yield more accurate characterizations 
of verb contexts and more accurate subcat-frame frequency estimates. If the other fea- 
tures are capturable via regular-expression rules, they can also be automatically detected 
in the manner described in the Section 2. For example, main verbs in relative clauses 
without relative pronouns may have a higher probability of having the feature "nnk", i.e., 
"(NP)(NP)(VERB)". 
More formally, let Y be a response variable taking as its value a subeat frame. Let 
X1, X2,..., XN be explanatory variables. Each Xi is associated with a feature expressed 
by one or a set of regular expressions. If a feature is expressed by one regular expression 
(R), the value of the feature is 1 if the occurrence of the verb matches R and 0 otherwise. 
If the feature is expressed by a set of regular expressions, its value is the label of the 
regular expression that the occurrence of the verb matches. The set of regular expressions 
in Table 2 can therefore be considered to characterize one explanatory variable whose 
value ranges from (NP+NP) to (REST). 
Now, we assume that a training corpus is available in which all verb tokens are given 
along with their subcat frames. By running our system on the training corpus, we can 
automatically generate a (N + 1)-dimensional contingency table. Table 3 is an example 
of a 2-dimensional contingency table with X = <OUTPUT OF SYSTEM> and Y = <REAL 
OCCURRENCES>. Using loglinear models \[Agresti, 1990\], we can derive fitted values of 
102 
each cell in the (N + 1)-dimensional contingency table. In the case of a saturated model, 
in which all kinds of interaction of variables up to (N + 1)-way interactions are included, 
the raw cell counts are the Maximum Likelihood solution. The fitted values are then used 
to estimate the subcat frame frequencies of a new corpus as follows. 
First, the system is run on the new corpus to obtain an N-dimensional contingency 
table. This table is considered to be an X1 - X2 ..... XN-marginal table. What we 
are aiming at is the Y margins that represent the real subcat frame frequencies of the 
new corpus. Assuming that the training corpus and the new corpus are homogeneous 
(e.g., reflecting similar sub-domains or samples of a common domain), we estimate the Y 
margins using Bayes theorem on the fitted values of the training corpus as follows: 
E(Y = k IX1 - X2 ..... XS~ marginal table of the new corpus) 
= ~...ZJ~,i~...iN+P(Y=klX1 =il,X~=i2,'",XN=iN) 
il i2 iN 
= ~'"~JV~,i,...i,~+ P(XI=il'X2=i~'''"XN=iNIY=k) P(Y=k) 
il i2 iN 
= EZ ,N÷ Ek, J~ili~...iNk' 
il 12 iN 
where ~iili~...i. + is the cell count of the X1 - X2 ..... XN marginal table of the new 
corpus obtained as the system output, and .h411i2...iN~ is the fitted value of the (N + 1)- 
dimensional contingency table of the training corpus based on a particular loglinear model. 
4.2 Lexical Heuristics 
The simplest application of the above method is to use a 2-way contingency table, as in 
Table 3. There are two possibilities to explore in constructing a 2-way contingency table. 
One is to sum up the cell counts of all the verbs in the training corpus and produce a 
single (large) general table. The other is to construct a table for each verb. Obviously 
the former approach is preferable if it works. Unfortunately, such a table is typically too 
general to be useful; the estimated frequencies based on it are less accurate than raw 
system output. This is because the sources of errors, viz., the distribution of off-diagonal 
cell counts of 2-way contingency tables, differ considerably from verb to verb. The latter 
approach is problematic if we have to make such a table for each domain. However, if we 
have a training corpus in one domain, and if the heuristics for each verb extracted from 
the training corpus are also applicable to other domains, the approach may work. 
To test the latter possibility, we constructed a contingency table for the verb from 
the test corpus described in the Section 3 that was most problematic (least accurately 
estimated) among the 33 verbs--"need". Note that we are using the test corpus described 
in the Section 3 as a training corpus here, because we already know both the measured 
frequency and the hand-judged frequency of "need" which are necessary to construct a 
contingency table. The total occurrence of this verb was 75. To smooth the table, 0.1 is 
added to all the cell counts. As new test corpora, we extracted another 300,000 words of 
tagged text from the WSJ corpus (labeled "W3") and also three sets of 300,000 words of 
tagged text from the Brown corpus (labeled "BI", "B2", and "B3"), as retagged under the 
103 
W3 
Measured 
By Hand 
Estimated 
NP+NP NP+CL NP+INF NP CL INF REST 
2.4 0.0 10.6 44.7 1.2 31.8 9.4 
0.0 0.0 0.0 69.4 0.0 30.6 0.0 
0.0 0.0 0.0 66.3 0.0 30.1 3.6 
Tot~ Occurrences:85 
B1 
Measured 
By Hand 
Estimated 
NP+NP NP+CL NP+INF NP CL INF REST 
1.8 0.9 7.9 38.6 1.8 14.9 34.2 
0.0 0.0 0.0 72.8 0.0 15.8 11.4 
0.0 0.0 0.0 76.6 0.0 14.4 9.1 
Tot~ Occurrences:ll4 
B2 
Measured 
By Hand 
Estimated 
NP+NP NP+CL NP+INF NP CL INF REST 
0.0 1.4 8.7 40.6 1.4 17.4 30.4 
0.0 0.0 0.0 73.9 0.0 18.8 7.2 
0.0 0.0 0.0 76.1 0.0 16.4 7.5 
TotalOccurrences:69 
B3 
Measured 
By Hand 
Estimated 
NP+NP NP+CL NP+INF NP CL INF REST 
3.3 0.0 1.7 30.0 3.3 31.7 30.0 
0.0 0.0 0.0 60.0 0.0 28.3 11.7 
0.0 0.0 0.0 61.4 0.0 29.8 8.8 
TotalOccurrences:60 
Table 5: Statistical Estimation (Unit = %) for the Verb "Need" 
Penn Treebank tagset. All the training and test corpora were reviewed--and judged--by 
hand. 
Table 5 gives the frequency distributions based on the system output, hand judge- 
ment, and statistical analysis. (As before, we take the hand judgement to be the gold 
standard, the actual frequency of a particular frame.) After the Y margins are statisti- 
cally estimated, the least estimated Y values less than 1.0 are truncated to 0. (These are 
considered to have appeared due to the smoothing.) 
In all of the test corpora, the method gives very accurate frequency distribution es- 
timates. Big gaps between the automatically-measured and manually-determined fre- 
quencies of "NP" and "REST" are shown to be substantially reduced through the use of 
statistical estimation. This result is especially encouraging because tile heuristics obtained 
in one domain are shown to be applicable to a considerably different domain. Further- 
more, by combining more feature sets and making use of multi-dimensional analysis, we 
can expect to obtain more accurate estimations. 
104 
5 Conclusion and Future Direction 
We have demonstrated that by combining syntactic and statistical analysis, the frequencies 
of verb-subcat frames can be estimated with high accuracy. Although the present system 
measures the frequencies of only six subcat frames, the method is general enough to be 
extended to many more frames. The traditional application of regular expressions as 
rules for deterministic processing has self-evident limitations since a linear grammar is 
not powerful enough to capture general linguistic phenomena. The statistical method we 
propose uses regular expressions as filters for detecting specific features of the occurrences 
of verbs and employs multi-dimensional analysis of the features based on loglinear models 
and Bayes Theorem. 
We expect that by identifying other useful syntactic features we can further improve 
the accuracy of the frequency estimation. Such features can be regarded as characterizing 
the syntactic context of the verbs, quite broadly. The features need not be linked to a 
local verb context. For example, a regular expression such as "w\['vex\]*k" can be used 
to find cases where the target verb is preceded by a relative pronoun such that there is 
no other finite verb or punctuation or sentence final period between the relative pronoun 
and the target verb. 
If the syntactic structure of a sentence can be predicted using only syntactic and lexical 
knowledge, we can hope to estimate the subcat frame of each occurrence of a verb using 
the context expressed by a set of features. We thus can aim to extend and refine this 
method for use with general probabilistic parsing of unrestricted text. 
6 Acknowledgements 
We thank Teddy Seidenfeid, Jeremy York, and Alex Franz for their comments and dis- 
cussions with us. We remain, of course, solely responsible for any errors or inadequacies 
in the paper. 

References 
\[Agresti, 1990\] A. Agresti. Categorical Data Analysis. New York, NY: John Wiley and 
Sons, 1990. 
\[Baker, 1979\] J. Baker. "Trainable Grammars for Speech Recognition". In D.II. Klatt 
and J.J. Wolf (eds.), Speech Communication Papers for the 97th Meeting of lhe Acoustic 
Society of America, 1979, pp. 547-550. 
\[Brent, 1'991\] M.R. Brent. "Automatic Acquisition of Subcategorization Frames from 
Untagged Text". Proceedings of the 29th Annual Meeting of the ACL, 1991. 
\[Brent and Berwick, 1991\] M.R. Brent and R.C. Berwick. "Automatic Acquisition of Sub- 
categorization Frames from Tagged Text". In Proceedings of the DARPA Speech and 
Natural Language Workshop, Morgan Kaufmann, 1991. 
\[Brill, 1992\] E. Brill. "A Simple Rule-Based Part of Speech Tagger". In Proceedings of 
the DARPA Speech and Natural Language Workshop, Morgan Kaufmann, 1992. 
\[Carroll and Charniak, 1992\] G. Carroll and E. Charniak. "Learning Probabilistic De- 
pendency Grammars from Labelled Text". In Working Notes of the Symposium on 
Probabilistic Approaches to Natural Language, AAAI Fall Symposium Series, 1992. 
\[Church, 1988\] K.W. Church. "A Stochastic Parts Program and Noun Phrase Parser 
for Unrestricted Text". In Proceedings of the Second Conference on Applied Natural 
Language Processing, 1988. 
\[de Marken, 1992\] C.G. de Mareken. "Parsing the LOB Corpus". In Proceedings of the 
gSth Annual Meeting of the ACL, 1990, pp. 243-251. 
\[Hornby, 1989\] A.S. Hornby, (ed.). Oxford Advanced Learner's Dictionary of Current 
English. Oxford, UK: Oxford University Press, 1989. 
\[Jelinek et al., 1990\] F. Jelinek, L.D. Lafferty, and R.L. Mercer. Basic Method of Proba- 
bilistic Context Free Grammars. Technical Report RC 16374 (72684), IBM, Yorktown 
Heights, NY 10598, 1990. 
\[Lauri and Young, 1990\] K. Lari and S.J. Young. "The Estimation of Stochastic Context- 
Free Grammars Using the Inside-Outside Algorithm". Computer Speech and Language, 
4, 1990, pp. 35-56. 
