We then integrate our features into full-scale dependency and constituent parsers .

We show relative error reductions of 7.0% over the second-order dependency parser of McDonald and Pereira (2006), 9.2% over the constituent parser of Petrov et al.

We next investigate the features that were given high weight by our learning algorithm (in the constituent parsing case).

For constituent parsing , we rerank the output of the Berkeley parser (Petrov et al., 2006).

To show end-to-end effectiveness, we incorporate our features into state-of-the-art dependency and constituent parsers .

For constituent parsing , we use a reranking framework (Charniak and Johnson, 2005; Collins and Koo, 2005; Collins, 2000) and show 9.2% relative error reduction over the Berkeley parser baseline.

We then add them to a constituent parser in a reranking approach.

4.2 Constituent Parsing

We also evaluate the utility of web-scale features on top of a state-of—the-art constituent parser — the Berkeley parser (Petrov et al., 2006), an unlexical-ized phrase-structure parser.

1For constituent parsers , there can be minor tree variations which can result in the same set of induced dependencies, but these are rare in comparison.

Statistical constituent parsers have gradually increased in accuracy over the past ten years.

Although syntax is becoming increasingly important for large-scale NLP applications, constituent parsing is slow—too slow to scale to the size of many potential consumer applications.

Deterministic algorithms for dependency parsing exist that can extract syntactic dependency structure very quickly (Nivre, 2008), but this approach is often undesirable as constituent parsers are more accurate and more adaptable to new domains (Petrov et al., 2010).

5 Constituent parsing with a cascade of chunkers

We use cascades of chunkers for full constituent parsing , building hierarchical constituents bottom-up.

We use the standard data sets for unsupervised constituency parsing research: for English, the Wall Street Journal subset of the Penn Treebank-3 (WSJ, Marcus et al.

This result suggests that improvements to low-level constituent prediction will ultimately lead to further gains in overall constituent parsing .

portantly, until recently it was the only unsupervised raw text constituent parser to produce results competitive with systems which use gold POS tags (Klein and Manning, 2002; Klein and Manning, 2004; Bod, 2006) — and the recent improved raw-text parsing results of Reichart and Rappoport (2010) make direct use of CCL without modification.

Constituency parses were produced using the Berkeley parser (Petrov and Klein, 2007).

In our complete model, which jointly extracts and compresses sentences, we choose whether or not to cut individual subtrees in the constituency parses

Assume a constituency parse 758 for every sentence 3.

While we use constituency parses rather than dependency parses, this model has similarities with the vine-growth model of Daume III (2006).