Chitrao, M. and Grishman, R. 1990. Statistical Parsing of Messages. In Proceedings of the June 1990 DARPA Speech and Natural Language Workshop. Hidden Valley, Pennsylvania.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....any new edges from being added to the parse chart. This behavior seriously degrades the robustness of a natural language system using this type of parser. A few recent works on probabilistic parsing have proposed algorithms and devices for efficient, robust chart parsing. Bobrow[3] and Chitrao[4] introduce agendabased probabilistic parsing algorithms, although neither describe their algorithms in detail. Both algo rithms use a strictly best first search. As both Chitrao and Magerman[12] observe, a best first search penalizes longer and more complex constituents (i.e. constituents which ....

Chitrao, M. and Grishman, R. 1990. Statistical Parsing of Messages. In Proceedings of the June 1990 DARPA Speech and Natural Language Workshop. Hidden Valley, Pennsylvania.

....an interpretation is defined as the product of the conditional probabilities of the rules which are applied in the derivation of the interpretation. The use of PCFG, in fact, resorts more to syntactic knowledge rather than to lexical knowledge, and its performance seems to be only moderately good (Chitrao and Grishman, 1990). There are also many methods proposed which more e#ectively make use of lexical knowledge. Collins (1997) proposes disambiguation through use of a generative probability model based on a lexicalized CFG (in fact, a restricted form of HPSG (Pollard and Sag, 1987) See also (Collins, 1996; ....

Chitrao, Mahesh V. and Ralph Grishman. 1990. Statistical parsing of messages. Proceedings of DARPA Speech and Natural Language Workshop, pages 263--266.

.... that determines the probability of its use wherever it is applicable (i.e. Stochastic CFG; SCFG (Booth and Thompson, 1973) or (2) associating different probabilities with a production depending on the particular nonterminal occurrence (on the RHS of a production) that is being rewritten (Chitrao and Grishman, 1990). In the latter case probabilities depend on the context (within a production) of the nonterminal being rewritten. In general, while there may be alternative ways of associating frequency information with grammars, the aim is always to provide a way of associating probabilities with alternatives ....

....grammar in this paper has much in common with recent approaches to statistical language modeling outside the LTAG tradition. Firstly, SLG integrates statistical preferences acquired from training data with an underlying wide coverage grammar, following an established line of research, for example (Chitrao and Grishman, 1990; Charniak and Carroll, 1994; Briscoe and Carroll, 1995) The paper discusses techniques for making preferences sensitive to context to avoid known shortcomings of the context independent probabilities of SCFG (see e.g. Briscoe and Carroll (1993) Secondly, SLG is lexical, since elementary trees ....

Chitrao, M. and Grishman, R. (1990). Statistical parsing of messages. In Proceedings of the Speech and Natural Language Workshop, pages 263-- 266, Hidden Valley, PA.

.... 148, 200] Tagging [10, 19, 28, 56, 57, 66, 90, 91, 124, 125, 126, 131, 138, 153, 163, 168, 188] HMMs [21, 22, 23, 24, 25, 49, 64, 67, 78, 115, 119, 155, 157, 160, 161] Search [156] The Inside Outside Algorithm [85, 86, 136, 137] Regression [20, 30, 29, 38, 41, 42, 45, 46, 154, 162] Partial Parsing [6, 7, 8, 9, 11, 37, 43, 47, 48, 51, 52, 53, 57, 58, 112, 65, 69, 70, 71, 72, 73, 74, 75, 76, 88, 100, 101, 102, 103, 104, 107, 110, 113, 114, 120, 121, 127, 132, 133, 134, 140, 142, 145, 147, 149, 152, 163, 164, 165, 166, 169, 178, 182, 186, 190, 191, 192, 194, 195, 196, 197] Grammatical Inference, Acquisition [1, 12, 13, 14, 15, 16, 32, 33, 39, 40, 55, 58, 79, 80, 83, 93, 94, 109, 111, 130, 167, 175, 179, 181, 184, 187, 189, 199] Mutual Information Parsing [98, 99, 146, 185] Prosody and Performance Structures [18, 26, 27, 31, 63, 92, 96, 97, 105, 106, 141, 151, 170, ....

M. Chitrao and R. Grishman. Statistical parsing of messages. In Proceedings of DARPA Speech and Natural Language Processing. Morgan Kaufman: New York, 1990.

....some sentence (n = 1; m = 3000) Increasing n leads to more rules being learnt and hence increases the plausibility of parses. The motivation for the edge limit follows from others who suggest that ungrammaticality might be related to an excessive number of edges being generated for some string [36, 37, 12]. In effect, the chart parser spends a lot of time fruitlessly searching for parses that may not exist. Resource bounds make learning incomplete, but if they are kept constant across all the learning configurations, any incompleteness effects will be factored out. However, incompleteness does mean ....

Mahesh V. Chitrao and Ralph Grishman. Statistical Parsing of Messages. In AAAI-92 Workshop Program: Statistically-Based NLP Techniques, San Jose, California, pages 263--266, 1992.

.... and hence increases the likelihood of finding highly plausible parses (i.e. those that match the benchmark parses exactly) The motivation for the edge limit follows from others who suggest that ungrammaticality might be related to an excessive number of edges being generated for some string [74, 75, 20]. In effect, the chart parser spends a lot of time fruitlessly searching for parses that may not exist. Resource bounds make learning incomplete, but if they are kept constant across all the learning configurations, any incompleteness effects will be factored out. However, incompleteness does mean ....

Mahesh V. Chitrao and Ralph Grishman. Statistical Parsing of Messages. In AAAI-92 Workshop Program: Statistically-Based NLP Techniques, San Jose, California, pages 263--266, 1992.

....sentences; using this grammar, sentences in this length range have produced up to 130,000 edges. Figure 1 shows a graph of non 0 E, that is, the percent of nonzero length edges needed to get 95 of the probability mass, for each sentence length. 5 Previous work Bobrow[1] and Chitrao and Grishman[5] introduced statistical agenda based parsing techniques. Chitrao and Grishman implemented a best first probabilistic parser and noted the parser s tendency to prefer shorter constituents. They proposed a heuristic solution of penalizing shorter constituents by a fixed amount per word. Miller and ....

Chitrao, Mahesh V. and Ralph Grishman (1990). Statistical Parsing of Messages. DARPA Speech and Language Workshop, 263-266.

....with the goal of nding a good parse (or the best parse) in the space of all possible parses. As such, we can build a parser which uses statistics to guide the search, thus speeding up the parsing process. In Best rst probabilistic parsing, introduced by Bobrow [1] and Chitrao and Grishman [5], partial parses are ranked according to some probabilistic gure of merit (FOM) and higher ranked subparses are given priority in the search that is, the parser preferentially works to expand subparses with higher FOMs. Naturally, the speed of a best rst parser and the correctness of its ....

Mahesh V. Chitrao and Ralph Grishman. Statistical parsing of messages. In DARPA Speech and Language Workshop, pages 263-266, 1990.

.... (Within LUST, there are provisions for using the IDLP grammar described in (Gazdar, Klein, Pullum, Sag, 1985) which uses some unordered rules unlike those used in more traditional studies of syntax and is attracting increasing interest in the linguistics community (Briscoe Carroll, 1993; Chitrao Grisham, 1990)) For the purposes of the LUST system, each gene contains a set of syntactic rules (a constant number of rules for each possible part of speech on the left hand side of a syntactic rule) Each non terminal symbol on the left hand side of a rule could have different rules describing its direct ....

Chitrao, M. V., & Grisham, R. (1990). Statistical parsing of messages. In Speech and Natural Language, pp. 263--266 Hidden Valley, PA.

....the ff and p(t 0;n ) terms in the straight fi figure above is that inside probability alone tends to prefer shorter constituents to longer ones, as the inside probability of a longer constituent involves the product of more probabilities. This can result in a thrashing effect as noted in [8], where the system parses short constituents, even very low probability ones, while avoiding combining them into longer constituents. To avoid thrashing, some technique is used to normalize the inside probability for use as a figure of merit. One approach is to take the geometric mean of the ....

....in [6] and the parser in that paper is a best first parser using the boundary trigram figure of merit. The literature shows many implementations of best first parsing, but none of the previous work shares our goal of explicitly comparing figures of merit. Bobrow [1] and Chitrao and Grishman [8] introduced statistical agendabased parsing techniques. Chitrao and Grishman implemented a best first probabilistic parser and noted the parser s tendency to prefer shorter constituents. They proposed a heuristic solution of penalizing shorter constituents by a fixed amount per word. Miller and ....

Mahesh V. Chitrao and Ralph Grishman. Statistical parsing of messages. In DARPA Speech and Language Workshop, pages 263--266, 1990.

.... called ARPA) DARPA 1989a, 1989b, 1990, 1991, 1992; ARPA, 1993, 1994) However, the book s bibliography fails to cite any papers from any of these workshops, many of which were important in the development of statistical NLP (e.g. Church et al. 1989, Gale and Church 1990, Brill et al. 1990, Chitrao and Grishman 1990, Magerman and Marcus 1991, Black et al. 1992a, 1992b, Brill 1992, and Lau 1993) Of the 44 bibliography entries, only two papers from Computational Linguistics are mentioned, omitting papers such as Brownet al... 1990, Seneff 1992 and Hindle and Rooth 1993footnoteHindle and Rooth 1993 is an ....

Chitrao, Mahesh and Grishman, Ralph (1990). "Statistical parsing of messages." Speech and Natural Language: Proceedings of a Workshop Held at Hidden Valley, Pennsylvania. Morgan Kaufmann Publishers, Inc., San Francisco, California.

.... grammar Using a sample of 260 sentence from the MUG 3 cor pus (disjoint from the set used here for evaluation) we computed the probability of each production in our context free grammar using an iterative unsuper vised training scheme similar to the inside outside al gorithm [Fujisaki, 1984; Chitrao and Grishman, 1990; Chitrao, 1990] We then used the logarithms of the probabilities as penalties in applying the productions during our analysis of the evaluation corpus. We used these statistical weights by themselves and in combination with closest attachment. Note that statistical weighting by itself is not ....

.... sentence from the MUG 3 cor pus (disjoint from the set used here for evaluation) we computed the probability of each production in our context free grammar using an iterative unsuper vised training scheme similar to the inside outside al gorithm [Fujisaki, 1984; Chitrao and Grishman, 1990; Chitrao, 1990]. We then used the logarithms of the probabilities as penalties in applying the productions during our analysis of the evaluation corpus. We used these statistical weights by themselves and in combination with closest attachment. Note that statistical weighting by itself is not particularly ....

Mahesh Chitrao and Ralph Grishman. Statistical parsing of messages. In Proceedings of the Speech and Natural Language Workshop, pages 263-266, Hidden Valley, PA, June 1990. Morgan Kaufmann.

....broader coverage than we might with a hand constructed grammar. Furthermore, experiments over the past few years have shown the benefits of using probabilistic information in parsing, and the large corpus allows us to train the probabilities of a grammar [Fujisaki 84] Garside and Leech 85] Chitrao and Grishman 90] Magerman and Weir 92] Black et al. 93] Bod 93] Magerman 95] Collins 96] A number of recent parsing experiments have also indicated that grammars whose production probabilities are dependent on the context can be more effective than context free grammars in selecting a correct parse. This ....

....than context free grammars in selecting a correct parse. This context sensitivity can be acquired easily using a large corpus, whereas human ability to compute such information is obviously limited. There have been several attempts to build context dependent grammars based on large corpora [Chitrao and Grishman 90] Simmons and Yu 91] Magerman and Weir 92] Schabes and Waters 93] Black et al. 93] Bod 93] Magerman 95] As is evident from the two lists of citations, there has been considerable research involving both probabilistic grammar based on syntactically bracketed corpora and context sensitivity. ....

[Article contains additional citation context not shown here]

Mahesh Chitrao, Ralph Grishman: "Statistical Parsing of Message" Proceeding of the June DARPA Speech and Natural Language workshop. Hidden Valley, Pennsylvania. (1990)

No context found.

Chitrao, M.V. k Grishman, R. (1990). Statistical Parsing of Messages. Proceedings of the DARPA Speech and Natural Language Workshop.

No context found.

Mahesh V. Chitrao and Ralph Grishman. 1990. Statistical parsing of messages. In DARPA Speech and Language Workshop, pages 263-266.

No context found.

Chitrao, M. and Grishman, R. 1990. Statistical Parsing of Messages. In Proceedings of the June 1990 DARPA Speech and Natural Language Workshop. Hidden Valley, Pennsylvania.

No context found.

Mahesh V. Chitrao and Ralph Grishman. 1990. Statistical parsing of messages. In DARPA Speech and Language Workshop, pages 263--266.

No context found.

Mahesh V. Chitrao and Ralph Grishman. 1990. Statistical parsing of messages. In DARPA Speech and Language Workshop, pages 263{ 266.

No context found.

Mahesh V. Chitrao & Ralph Grishman (1990). Statistical parsing of messages, in Proceedings of the June 1990 DARPA Speech and Natural Language Workshop, San Mateo: Morgan Kaufmann, 263-266.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC