PropBank - Semantic Role Labeling and Proposition Bank

In the mid-2000s, as statistical natural language processing matured and semantic role labeling emerged as a critical task, researchers at the University of Pennsylvania undertook an ambitious project that would transform how computational linguistics approached semantic analysis. PropBank, the Proposition Bank project developed under the leadership of Martha Palmer, Daniel Gildea, and Paul Kingsbury, represented a crucial advance by adding semantic role labels to the existing Penn Treebank syntactic annotations. Launched in 2005, PropBank created the first large-scale semantic annotation resource that was directly compatible with a major syntactic treebank, enabling statistical systems to learn "who did what to whom" in natural language sentences while leveraging the rich syntactic structure already present in the Penn Treebank.

The project emerged at a critical moment. By 2005, the Penn Treebank had become the standard resource for training statistical parsers, but syntactic structure alone couldn't capture the semantic content crucial for natural language understanding. FrameNet, released in 1998, had demonstrated the value of semantic role annotation, but its frame-based approach with named roles like Buyer and Seller covered only a limited vocabulary. PropBank aimed to provide broader coverage by using a simpler, more scalable annotation scheme based on numbered arguments (Arg0, Arg1, Arg2, etc.) rather than named frame elements. This design choice enabled PropBank to annotate substantially more lexical units while maintaining compatibility with existing Penn Treebank trees, creating a comprehensive resource that combined syntactic and semantic information.

PropBank's systematic annotation of predicate-argument structure provided the foundation for semantic role labeling as a standard NLP task. The project produced semantically annotated text that made it possible to train and evaluate statistical systems for automatically identifying semantic roles, establishing semantic role labeling as a central benchmark for evaluating semantic analysis systems. The availability of large amounts of training data enabled machine learning approaches to semantic role labeling that could achieve high accuracy on realistic text, demonstrating that large-scale semantic annotation was both feasible and valuable for computational linguistics.

The success of PropBank established important principles for semantic annotation that would influence subsequent projects. Its compatibility with existing syntactic annotation, systematic use of frame files to define verb-specific roles, and emphasis on annotation consistency became standard practices in semantic resource development. PropBank's influence extended to the CoNLL shared tasks on semantic role labeling, which became the standard evaluation framework for semantic analysis systems, and to subsequent annotation projects like OntoNotes. The project demonstrated that semantic annotation could scale to cover substantial portions of vocabulary while maintaining quality, establishing PropBank as a foundational resource for computational semantics.

By 2005, statistical NLP systems had achieved substantial progress in syntactic parsing, but they faced a fundamental limitation: understanding who did what to whom required semantic knowledge that syntactic structure alone couldn't provide. Consider the sentences "John gave Mary a book" and "Mary received a book from John." These sentences have different syntactic structures but express the same semantic content involving a transfer of possession. Syntactic parsers could identify the grammatical relationships, but they couldn't capture that John is the agent of giving, Mary is the recipient, and the book is the theme being transferred. This semantic information is crucial for natural language understanding tasks like information extraction, question answering, and machine translation.

The problem extended beyond individual sentences. Different syntactic realizations could express the same semantic roles. In "The company bought the subsidiary," the subject "The company" is the buyer. But in "The subsidiary was bought by the company," the buyer appears in a prepositional phrase, while the object becomes the subject. Systems needed to recognize that despite different syntactic structures, the semantic roles remain the same. Traditional resources couldn't help with this. Syntactic parsers provided tree structures but not semantic roles. WordNet provided word relationships but not event structure. The gap between syntactic analysis and semantic understanding constrained NLP systems' capabilities.

This limitation became increasingly problematic as systems attempted more sophisticated applications. Information extraction systems needed to identify entities and their relationships, but without understanding semantic roles, they struggled to determine whether a person mentioned in a sentence was an agent, a patient, a beneficiary, or some other participant. Question answering systems needed to understand who performed actions, on what objects, for what purposes, but existing resources provided no framework for encoding this information systematically. Machine translation systems needed to preserve semantic relationships across languages, but word-level resources couldn't capture the event structures that verbs and other predicates evoked.

FrameNet, released in 1998, had demonstrated that semantic role annotation was valuable and feasible. FrameNet's frame-based approach with named roles like Buyer, Seller, and Goods provided detailed semantic structures, and it showed that large-scale semantic annotation of corpus data could produce useful resources. However, FrameNet's annotation process was labor-intensive, requiring detailed frame definitions for each lexical unit. By 2005, FrameNet had annotated around 13,000 lexical units, but English vocabulary includes hundreds of thousands of words. The coverage gap limited FrameNet's usefulness for unrestricted text processing, and its frame-based approach required substantial annotation effort for each new lexical unit.

Researchers at the University of Pennsylvania recognized that a different approach could provide broader coverage more efficiently. Rather than creating detailed named frames for each verb, a simpler scheme using numbered arguments could scale more effectively while still capturing the essential semantic information needed for semantic role labeling. This scheme could build directly on the Penn Treebank's existing syntactic annotations, avoiding the need to create a separate resource from scratch. The challenge was designing an annotation scheme that was simple enough to scale broadly but precise enough to capture meaningful semantic distinctions. PropBank addressed this challenge by combining verb-specific frame files with a general numbering scheme for arguments.

PropBank addressed this problem by creating a semantic annotation layer that worked directly with the existing Penn Treebank syntactic trees. The solution had two key components: numbered arguments that could scale across different verbs, and verb-specific frame files that defined what each numbered argument represented for particular verbs. This design enabled PropBank to provide broader coverage than FrameNet while maintaining compatibility with the most widely used syntactic resource in computational linguistics.

PropBank's core innovation was using numbered arguments (Arg0, Arg1, Arg2, etc.) rather than named semantic roles. For verbs, these arguments had general semantic interpretations:

• Arg0: Typically the agent or proto-agent (the entity that performs the action)
• Arg1: Typically the patient, theme, or proto-patient (the entity affected by the action)
• Arg2: Typically the beneficiary, recipient, or instrument
• Arg3: Often the start point, benefactive, or attribute
• Arg4: Often the end point

These general interpretations provided consistency across verbs, but PropBank's real power came from verb-specific frame files that defined precisely what each numbered argument meant for each verb. For the verb "give," the frame file might specify that Arg0 is the giver, Arg1 is what is given, and Arg2 is the recipient. For "buy," Arg0 would be the buyer, Arg1 would be what is bought, and Arg2 might be the seller or source. This combination of general numbering and verb-specific definitions enabled PropBank to scale while maintaining semantic precision.

The numbered argument scheme had several advantages. It was simpler to annotate than creating detailed named frames, enabling faster annotation and broader coverage. It was compatible with the Penn Treebank's tree structures, allowing annotators to mark semantic roles directly on existing syntactic nodes. It provided a framework that machine learning systems could learn from, with consistent argument numbering patterns across verbs. And it allowed verb-specific distinctions through frame files, avoiding the one-size-fits-all problem that purely general schemes might face.

Each verb in PropBank was associated with a frame file that defined the possible semantic roles for that verb. These frame files specified what each numbered argument (Arg0, Arg1, Arg2, etc.) represented for that particular verb, ensuring consistency in annotation across different occurrences of the same verb. Frame files also documented the syntactic realizations typical for each argument, such as whether Arg0 typically appears as the subject or whether Arg2 typically appears with a preposition.

Consider the verb "break." Its frame file might specify that Arg0 is the agent that causes the breaking (typically the subject), Arg1 is what gets broken (typically the object), and Arg2 might be the instrument used (typically a prepositional phrase with "with"). The frame file captures both the semantic roles and their typical syntactic realizations, providing a complete picture of how this verb structures events. When annotators encountered "John broke the window with a hammer," they could consistently mark John as Arg0 (the agent), "the window" as Arg1 (what was broken), and "a hammer" as Arg2 (the instrument).

Frame files also handled verb alternations, where the same verb can appear in different syntactic frames while maintaining similar semantics. The verb "load" can appear as "John loaded the truck with boxes" (where Arg0 is the agent, Arg1 is the location, and Arg2 is what is loaded) or "John loaded boxes onto the truck" (where Arg0 is the agent, Arg1 is what is loaded, and Arg2 is the location). The frame file documents both patterns, showing how the same verb can realize its arguments differently while maintaining consistent semantic roles.

The PropBank annotation process involved several systematic steps. First, annotators identified the main predicate of each sentence, typically a verb. Then, they consulted the frame file for that verb to determine what semantic roles were possible. Next, they identified the arguments of the predicate in the Penn Treebank tree and determined which numbered arguments they represented based on the frame file. The annotation also included information about the syntactic realization of each argument, such as whether it was a noun phrase, prepositional phrase, or other syntactic category.

This dual annotation of semantic roles and syntactic realization made PropBank useful for both semantic and syntactic analysis. Systems could use the semantic role labels directly for semantic understanding tasks, or they could study the relationship between syntax and semantics by examining how semantic roles were realized syntactically. This combination proved valuable for research on argument alternations, voice constructions, and other phenomena where syntactic and semantic structures interact.

The annotation scheme was designed to be compatible with the existing Penn Treebank syntactic annotation. Semantic roles were marked on the same tree structures that syntactic annotation used, allowing the two levels of annotation to be used together seamlessly. This compatibility meant that researchers didn't need to choose between syntactic and semantic information: they could leverage both, leading to more sophisticated natural language understanding systems that combined syntactic and semantic analysis.

PropBank's simpler annotation scheme enabled broader coverage than FrameNet. By 2010, PropBank had annotated over 1 million words of text with semantic roles, covering substantially more lexical units than FrameNet's detailed frame annotations. This broader coverage made PropBank more useful for applications that needed to process unrestricted text, where encountering unannotated lexical units was common. The resource became the standard for semantic role labeling tasks precisely because of this coverage advantage.

The coverage advantage came from PropBank's efficient annotation process. Rather than creating detailed frame definitions for each lexical unit, annotators could use the general numbered argument scheme with verb-specific frame files. This process was faster than FrameNet's frame creation, enabling PropBank to annotate more text in less time. The trade-off was less detailed semantic structures than FrameNet provided, but for many applications, the broader coverage was more valuable than the additional detail.

PropBank's release in 2005 had immediate and lasting impact on natural language processing research. The resource enabled semantic role labeling as a standard NLP task, providing both the framework for specifying roles and the annotated training data needed for statistical systems. Within a few years, PropBank had become the foundation for semantic role labeling evaluation through the CoNLL shared tasks, establishing it as a crucial resource for computational semantics.

The most direct application of PropBank was semantic role labeling (SRL), the task of automatically identifying which phrases in a sentence fill which semantic roles. PropBank provided both a framework for specifying roles and substantial annotated training data, making SRL a practical task for statistical NLP systems. Early SRL systems trained on PropBank data could learn patterns like: when "give" appears as the main verb, the subject typically fills Arg0 (the giver), the direct object typically fills Arg1 (what is given), and a prepositional phrase with "to" typically fills Arg2 (the recipient).

These patterns enabled systems to label semantic roles in new text automatically. By the late 2000s, semantic role labeling had become a standard NLP task with evaluation benchmarks and shared tasks. The CoNLL-2004 and CoNLL-2005 shared tasks on semantic role labeling were based on PropBank data, establishing PropBank as the standard resource for SRL evaluation. These shared tasks helped establish semantic role labeling as a core NLP capability, demonstrating that automatic semantic analysis was achievable at scale.

Modern semantic role labeling systems, including neural approaches, continue to use PropBank annotations as training data. The task remains important for applications that need to understand event structure, such as information extraction, question answering, and text summarization. PropBank's role in establishing SRL as a standard task represents one of its most significant contributions to computational linguistics.

Information extraction systems benefited substantially from PropBank's structured semantic annotations. Traditional information extraction focused on identifying entities and binary relations, but struggled with events that involved multiple participants with specific roles. PropBank provided a framework for representing these events systematically, enabling systems to extract structured information about who did what to whom, where, when, and why.

Consider extracting information from news articles about corporate events. A sentence like "Acme Corp acquired TechStart Inc for $50 million in 2005" requires identifying the acquirer (Acme Corp), the target (TechStart Inc), the price ($50 million), and the date (2005). PropBank's annotations enable systems to recognize that "acquired" is the predicate, with the subject as Arg0 (the acquirer), the object as Arg1 (the target), and prepositional phrases providing additional roles. Systems trained on PropBank data could learn these patterns and apply them to extract structured information from unstructured text automatically.

PropBank's compatibility with syntactic trees also enabled more sophisticated extraction systems that combined syntactic and semantic analysis. These systems could use syntactic information to identify argument boundaries more accurately, then use semantic role labels to determine what type of information each argument represented. This combination proved more accurate than approaches that used only syntactic or only semantic information.

Question answering systems found PropBank invaluable for understanding what questions ask and where answers might be found. Consider the question "Who gave the book to Mary?" This question asks for the Arg0 (the giver) in a "give" event where Arg1 is "the book" and Arg2 is "Mary." A question answering system using PropBank would recognize this structure, then search text for sentences where "give" (or related predicates) appears with matching arguments, extract the phrase filling the Arg0 role, and return it as the answer.

PropBank also enabled more sophisticated question answering where questions involve complex events with multiple participants. "Who sold what to whom for how much?" requires understanding multiple roles in a "sell" event. Systems using PropBank could parse such questions into semantic role structures, then match them against similarly structured representations of text passages. This semantic matching proved more accurate than approaches that didn't use structured semantic representations, enabling question answering systems to understand and answer questions that required semantic understanding beyond simple keyword matching.

Machine translation systems, particularly statistical machine translation systems, used PropBank to preserve semantic roles across languages. Translation systems need to ensure that when they translate a sentence, the semantic relationships between participants remain consistent. If a source sentence has "John gave Mary a book," where John is Arg0 (the giver), Mary is Arg2 (the recipient), and the book is Arg1 (what is given), the translation should preserve these roles even if the target language expresses them differently syntactically.

PropBank provided a language-independent representation of these roles. Translation systems could map source language sentences to PropBank-style semantic structures, then generate target language sentences from those structures, ensuring that roles were preserved even when syntax differed. This approach was particularly valuable for translation between languages with different syntactic structures, where direct word alignment might lose semantic information. Systems that preserved semantic roles produced more accurate translations, especially for sentences involving complex event structures.

PropBank's compatibility with the Penn Treebank made it valuable for research that combined syntactic and semantic analysis. Researchers could study how semantic roles were realized syntactically, examining phenomena like argument alternations, voice constructions, and control structures. This research advanced understanding of the syntax-semantics interface and enabled development of systems that leveraged both types of information.

PropBank also influenced the development of other semantic annotation projects. The OntoNotes project, which created multi-layered annotations including syntax, semantics, and discourse, incorporated PropBank-style semantic role annotations. This integration demonstrated PropBank's influence on subsequent annotation efforts and its role as a foundational resource for computational semantics. The combination of PropBank with other annotation layers in OntoNotes created a comprehensive resource for multi-level language analysis.

Despite PropBank's significant contributions, it faced several important limitations that constrained its practical applications and highlighted challenges inherent in semantic annotation at scale. These limitations reflected both the trade-offs made in PropBank's design choices and the inherent difficulty of comprehensively annotating semantic knowledge.

PropBank's most obvious limitation was its coverage. Even after substantial annotation efforts, PropBank covered only a fraction of the verbs and predicates that appear in real text. Many common verbs lacked frame files, limiting PropBank's usefulness for unrestricted text processing. The annotation process, while more efficient than FrameNet's, still required creating frame files for each verb and annotating sentences, a labor-intensive process that didn't scale easily to cover the full vocabulary.

The coverage problem was exacerbated by the fact that PropBank focused primarily on verbal predicates. While verbs are central to event structure, many sentences involve nominalizations, adjectives, and other predicates that also have semantic roles. A sentence like "The acquisition of TechStart by Acme Corp" involves an event with roles, but "acquisition" is a noun, not a verb. PropBank's focus on verbs limited its coverage of these alternative predicate types, constraining its usefulness for some applications.

PropBank's use of verb-specific frame files created a tension between verb-specific distinctions and general role patterns. Some researchers argued that PropBank's verb-specific approach missed generalizations that could be captured with more general role labels. For example, many verbs have similar argument structures that could be grouped together, but PropBank's frame files treated each verb separately, potentially missing these patterns.

The verb-specific approach also created challenges for handling novel or rare verbs. If a system encountered a verb that didn't have a frame file in PropBank, it couldn't use PropBank's role definitions. While systems could fall back to general interpretations of Arg0, Arg1, etc., these interpretations were less precise than verb-specific definitions. This limitation affected PropBank's usefulness for domains with specialized vocabulary or for processing text with many rare verbs.

PropBank's annotations, while valuable, reflected the inherent subjectivity in semantic annotation. Different annotators sometimes disagreed about which phrases filled which roles, especially for arguments that were less clearly defined or for verbs with multiple possible argument structures. This annotation inconsistency affected the quality of training data derived from PropBank. Machine learning systems trained on inconsistent annotations learned inconsistent patterns, reducing their accuracy.

The annotation process also faced challenges with ambiguity and underspecification in natural language. Consider "John broke the window." This sentence clearly involves a breaking event, but many semantic roles are unspecified: What instrument was used? When did it happen? Why did John break it? PropBank's annotations marked only the roles that were explicitly realized in the sentence, but this meant that many annotations were incomplete. Systems couldn't always distinguish between roles that were unrealized but implied versus roles that simply weren't relevant to the event.

PropBank's use of numbered arguments rather than named roles was a design choice that enabled scalability but created limitations. Numbered arguments like Arg0 and Arg1 are less interpretable than named roles like Agent and Patient. Researchers and system developers had to consult frame files to understand what each argument represented, adding complexity to using PropBank annotations.

The numbered argument scheme also created challenges for applications that needed human-readable semantic representations. While Arg0 might consistently represent the agent across many verbs, the label "Arg0" doesn't convey this meaning directly. Named roles like Agent are more interpretable, but PropBank's design prioritized scalability over interpretability. This trade-off limited PropBank's usefulness for some applications that needed more transparent semantic representations.

PropBank's annotations were based primarily on general-purpose corpora like the Wall Street Journal texts used in the Penn Treebank. Many applications needed semantic role knowledge in specialized domains like medicine, law, or finance. PropBank's general-purpose annotations didn't always capture domain-specific event structures. A medical frame for diagnosis might require different elements than a general-purpose frame for observation, but creating domain-specific PropBank annotations required new annotation efforts.

The domain specificity limitation affected PropBank's usefulness as applications moved into specialized domains where comprehensive semantic knowledge was most needed. While PropBank provided a foundation, domain-specific applications often needed additional annotation or adaptation of PropBank's general-purpose frame files to handle domain-specific predicates and events.

Despite these limitations, PropBank remained an influential resource because it addressed a fundamental need: providing broad-coverage semantic annotations that were compatible with existing syntactic resources. Even if coverage was incomplete and annotation was subjective, PropBank provided a foundation that researchers could build upon. The limitations highlighted the challenges of semantic resource development at scale, but didn't diminish PropBank's contribution to establishing semantic role labeling as a practical NLP capability.

PropBank's legacy extends far beyond its direct applications in semantic role labeling and information extraction. The resource established semantic role labeling as a standard NLP task, creating evaluation frameworks and training data that continue to influence modern NLP systems. Even as neural methods have transformed NLP, PropBank's approach to semantic annotation continues to shape how systems understand meaning and event structure.

PropBank's most direct legacy is the CoNLL shared tasks on semantic role labeling, which became the standard evaluation framework for semantic analysis systems. The CoNLL-2004 and CoNLL-2005 shared tasks, both based on PropBank data, established semantic role labeling as a core NLP benchmark. These shared tasks provided standardized evaluation procedures and enabled comparison across different SRL systems, advancing the state of the art in semantic role labeling.

The CoNLL shared tasks also influenced how semantic role labeling systems are designed and evaluated. They established evaluation metrics, standardized data formats, and created test sets that enabled fair comparison across systems. This standardization proved crucial for advancing the field, enabling researchers to build on each other's work and measure progress systematically. The shared tasks' influence extends to modern SRL evaluation, where PropBank-style annotations remain the standard.

Modern semantic role labeling systems, including neural approaches, continue to use PropBank annotations as training data. Deep learning models have achieved substantial improvements in SRL accuracy, but they still rely on PropBank's annotated data to learn semantic role patterns. The combination of PropBank's structured annotations with neural methods has enabled SRL systems that achieve high accuracy on realistic text, demonstrating PropBank's continuing relevance.

Recent neural SRL systems have moved beyond simply using PropBank annotations to incorporating semantic role structure into their architectures. Some systems explicitly model argument structures, learning representations that capture how predicates relate to their arguments. These systems leverage PropBank's frame files to structure their predictions, showing how PropBank's approach to semantic representation continues to guide modern approaches to semantic understanding. PropBank's framework has proven compatible with neural methods, demonstrating its lasting conceptual value.

Abstract Meaning Representation (AMR), developed in the 2010s, represents sentences as directed acyclic graphs capturing semantic structure. While AMR uses different primitives than PropBank, being rooted in linguistic predicates rather than numbered arguments, it shares PropBank's emphasis on event structure and semantic roles. AMR's representation of events with roles closely parallels PropBank's predicate-argument structures, showing how PropBank's approach to semantic representation has influenced subsequent schemes.

Modern AMR parsers sometimes use PropBank information during parsing, demonstrating PropBank's continuing relevance even as new representation schemes emerge. The predicate-argument approach, which represents events with structured roles, has proven robust across multiple formalisms, suggesting that PropBank captured something fundamental about how meaning should be represented computationally. AMR and PropBank represent different points on a spectrum of semantic representation, but both build on the insight that event structure is crucial for understanding meaning.

Modern knowledge graphs and event extraction systems build directly on PropBank's insights about event structure. Knowledge graphs represent events with predicates and arguments, maintaining structured representations similar to PropBank's predicate-argument structures. When a knowledge graph represents "John gave Mary a book" as an event with roles for agent, recipient, and theme, it's using the same conceptual structure that PropBank formalized, just with different notation.

PropBank's influence appears most clearly in event extraction systems that need to identify and structure events in text. These systems often use predicate-argument structures to represent events, with roles determined by the semantic type of event rather than purely syntactic positions. Modern event extraction, whether rule-based or neural, benefits from PropBank's demonstration that event structure is crucial for understanding meaning. The task of identifying who did what to whom, where, and when remains central to NLP, and PropBank provided the framework for addressing it systematically.

Interestingly, recent research suggests that large neural language models implicitly learn semantic role structures during training. Studies have shown that models like BERT and GPT can predict semantic roles and predicate-argument structures even without explicit PropBank-style training. This suggests that semantic role patterns are fundamental to language understanding, patterns that neural models discover through exposure to text. PropBank's explicit semantic structures might be providing a lens for understanding what neural models learn implicitly.

Some researchers are now combining explicit semantic resources like PropBank with neural language models, using semantic structures to interpret and improve neural representations. This hybrid approach leverages both PropBank's structured knowledge and neural models' ability to learn from large amounts of text. Semantic role structure provides interpretability and explicit event understanding, while neural methods provide coverage and flexibility. This combination suggests that PropBank's approach to semantic representation will remain relevant even as neural methods continue to advance.

PropBank's compatibility with syntactic trees has enabled its integration into modern NLP pipelines that combine multiple levels of analysis. Systems can use syntactic parsers to identify argument boundaries, then use PropBank annotations to determine semantic roles, creating multi-level representations that leverage both syntactic and semantic information. This integration has proven valuable for applications that need comprehensive language understanding.

Modern NLP pipelines often incorporate PropBank-style semantic role labeling as a component, alongside syntactic parsing, named entity recognition, and other tasks. The combination of these components enables more sophisticated language understanding systems that can leverage multiple levels of linguistic structure. PropBank's role in these pipelines demonstrates its continuing practical value, even as new methods emerge for individual tasks.

PropBank's dual annotation of semantic roles and syntactic realization has enabled substantial research on the syntax-semantics interface. Researchers have used PropBank to study how semantic roles are realized syntactically, examining phenomena like argument alternations, voice constructions, and control structures. This research has advanced understanding of how syntax and semantics interact, providing insights that inform both linguistic theory and computational applications.

The research enabled by PropBank's annotations has influenced how modern systems model the relationship between syntax and semantics. Some neural parsers now explicitly model semantic roles alongside syntactic structure, creating unified representations that capture both levels of analysis. This integration builds on PropBank's demonstration that syntactic and semantic annotation can work together effectively.

Today, PropBank continues to be maintained and expanded, with ongoing annotation efforts and regular releases. The resource has grown to cover hundreds of thousands of words with semantic annotations, providing training data for modern SRL systems. But perhaps more importantly, PropBank's core insights have become foundational to modern NLP: events have structured roles, these roles can be annotated systematically, and this structure is essential for understanding meaning. Even as new methods emerge, these insights remain central to how systems understand language and meaning.

Ready to test your understanding of PropBank and semantic role labeling? Challenge yourself with these questions about how PropBank added semantic structure to syntactic trees and how it enabled semantic role labeling as a standard NLP task. Good luck!