Log In Sign Up

A Matter of Framing: The Impact of Linguistic Formalism on Probing Results

by   Ilia Kuznetsov, et al.

Deep pre-trained contextualized encoders like BERT (Delvin et al., 2019) demonstrate remarkable performance on a range of downstream tasks. A recent line of research in probing investigates the linguistic knowledge implicitly learned by these models during pre-training. While most work in probing operates on the task level, linguistic tasks are rarely uniform and can be represented in a variety of formalisms. Any linguistics-based probing study thereby inevitably commits to the formalism used to annotate the underlying data. Can the choice of formalism affect probing results? To investigate, we conduct an in-depth cross-formalism layer probing study in role semantics. We find linguistically meaningful differences in the encoding of semantic role- and proto-role information by BERT depending on the formalism and demonstrate that layer probing can detect subtle differences between the implementations of the same linguistic formalism. Our results suggest that linguistic formalism is an important dimension in probing studies, along with the commonly used cross-task and cross-lingual experimental settings.


A Simple Geometric Method for Cross-Lingual Linguistic Transformations with Pre-trained Autoencoders

Powerful sentence encoders trained for multiple languages are on the ris...

BERT Busters: Outlier LayerNorm Dimensions that Disrupt BERT

Multiple studies have shown that BERT is remarkably robust to pruning, y...

Cross-Lingual Ability of Multilingual BERT: An Empirical Study

Recent work has exhibited the surprising cross-lingual abilities of mult...

What does it mean to be language-agnostic? Probing multilingual sentence encoders for typological properties

Multilingual sentence encoders have seen much success in cross-lingual m...

Finding Concept-specific Biases in Form–Meaning Associations

This work presents an information-theoretic operationalisation of cross-...

Probing as Quantifying the Inductive Bias of Pre-trained Representations

Pre-trained contextual representations have led to dramatic performance ...

Not All Models Localize Linguistic Knowledge in the Same Place: A Layer-wise Probing on BERToids' Representations

Most of the recent works on probing representations have focused on BERT...

1 Introduction

The emergence of deep pre-trained contextualized encoders has had a major impact on the field of natural language processing. Boosted by the availability of general-purpose frameworks like AllenNLP

Gardner et al. (2018) and Transformers Wolf et al. (2019), pre-trained models like ELMO Peters et al. (2018) and BERT Devlin et al. (2019) have caused a shift towards simple architectures where a strong pre-trained encoder is paired with a shallow downstream model, often outperforming the intricate task-specific architectures of the past.

Figure 1: Intra-sentence similarity by layer of the multilingual BERT-base. Functional words are similar in , syntactic groups emerge at higher levels.

The versatility of pre-trained representations implies that they encode some aspects of general linguistic knowledge Reif et al. (2019). Indeed, even an informal inspection of layer-wise intra-sentence similarities (Fig. 1) suggests that these models capture elements of linguistic structure, and those differ depending on the layer of the model. A grounded investigation of these regularities allows to interpret the model’s behavior, design better pre-trained encoders and inform the downstream model development. Such investigation is the main subject of probing, and recent studies confirm that BERT implicitly captures many aspects of language use, lexical semantics and grammar Rogers et al. (2020).

Most probing studies use linguistics as a theoretical scaffolding and operate on a task level. However, there often exist multiple ways to represent the same linguistic task: for example, English dependency syntax can be encoded using a variety of formalisms, incl. Universal Schuster and Manning (2016), Stanford de Marneffe and Manning (2008) and CoNLL-2009 dependencies Hajič et al. (2009), all using different label sets and syntactic head attachment rules. Any probing study inevitably commits to the specific theoretical framework used to produce the underlying data. The differences between linguistic formalisms, however, can be substantial.

Can these differences affect the probing results? This question is intriguing for several reasons. Linguistic formalisms are well-documented, and if the choice of formalism indeed has an effect on probing, cross-formalism comparison will yield new insights into the linguistic knowledge obtained by contextualized encoders during pre-training. If, alternatively, the probing results remain stable despite substantial differences between formalisms, this prompts a further scrutiny of what the pre-trained encoders in fact encode. Finally, on the reverse side, cross-formalism probing might be used as a tool to empirically compare the formalisms and their language-specific implementations. To the best of our knowledge we are the first to explicitly address the influence of formalism on probing.

Ideally, the task chosen for a cross-formalism study should be encoded in multiple formalisms using the same textual data to rule out the influence of the domain and text type. While many linguistic corpora contain several layers of linguistic information, having the same textual data annotated with multiple formalisms for the same task is rare. We focus on role semantics – a family of shallow semantic formalisms at the interface between syntax and propositional semantics that assign roles to the participants of natural language utterances, determining who did what to whom, where, when etc. Decades of research in theoretical linguistics have produced a range of role-semantic frameworks that have been operationalized in NLP: syntax-driven PropBank Palmer et al. (2005), coarse-grained VerbNet Kipper-Schuler (2005), fine-grained FrameNet Baker et al. (1998), and, recently, decompositional Semantic Proto-Roles (SPR) Reisinger et al. (2015); White et al. (2016). The SemLink project Bonial et al. (2013) offers parallel annotation for PropBank, VerbNet and FrameNet for English. This allows us to isolate the object of our study: apart from the role-semantic labels, the underlying data and conditions for the three formalisms are identical. SR3DE Mújdricza-Maydt et al. (23-28) provides compatible annotation in three formalisms for German, enabling cross-lingual validation of our results. Combined, these factors make role semantics an ideal target for a cross-formalism probing study.

A solid body of evidence suggests that encoders like BERT capture syntactic and lexical-semantic properties, but only few studies have considered probing for predicate-level semantics Tenney et al. (2019); Kovaleva et al. (2019). To the best of our knowledge we are the first to conduct a cross-formalism probing study on role semantics, thereby contributing to the line of research on how and whether pre-trained BERT encodes higher-level semantic phenomena.


This work studies the effect of the linguistic formalism on probing results. We conduct cross-formalism experiments on PropBank, VerbNet and FrameNet role prediction in English and German, and show that the formalism can affect probing results in a linguistically meaningful way; in addition, we demonstrate that layer probing can detect subtle differences between implementations of the same formalism in different languages. On the technical side, we advance the recently introduced edge and layer probing framework Tenney et al. (2019); in particular, we introduce anchor tasks - an analytical tool inspired by feature-based systems that allows deeper qualitative insights into the pre-trained models’ behavior. Finally, advancing the current knowledge about the encoding of predicate semantics in BERT, we perform a fine-grained semantic proto-role probing study and demonstrate that semantic proto-role properties can be extracted from pre-trained BERT, contrary to the existing reports. Our results suggest that along with task and language, linguistic formalism is an important dimension to be accounted for in probing research.

2 Related Work

2.1 BERT as Encoder

BERT is a Transformer Vaswani et al. (2017) encoder pre-trained by jointly optimizing two unsupervised objectives: masked language model and next sentence prediction. It uses WordPiece (WP, Wu et al. (2016)) subword tokens along with positional embeddings as input, and gradually constructs sentence representations by applying token-level self-attention pooling over a stack of layers . The result of BERT encoding is a layer-wise representation of the input wordpiece tokens with higher layers representing higher-level abstractions over the input sequence. Thanks to the joint pre-training objective, BERT can encode words and sentences in a unified fashion: the encoding of a sentence or a sentence pair is stored in a special token [CLS].

To facilitate multilingual experiments, we use the multilingual BERT-base (mBERT) published by Devlin et al. (2019). Although several recent encoders have outperformed BERT on benchmarks Liu et al. (2019); Lan et al. (2019); Raffel et al. (2019), we use the original BERT architecture, since it allows us to inherit the probing methodology and to build upon the related findings.

2.2 Probing

Due to space limitations we omit high-level discussions on benchmarking Wang et al. (2018) and sentence-level probing Conneau et al. (2018), and focus on the recent findings related to the representation of linguistic structure in BERT. Surface-level information generally tends to be represented in the lower layers of deep encoders, while higher layers store hierarchical and semantic information Belinkov et al. (2017); Lin et al. (2019). Tenney et al. (2019) show that the abstraction strategy applied by the English pre-trained BERT encoder follows the order of the classical NLP pipeline. Strengthening the claim about linguistic capabilities of BERT, Hewitt and Manning (2019) demonstrate that BERT implicitly learns syntax, and Reif et al. (2019) show that it encodes fine-grained lexical-semantic distinctions. Rogers et al. (2020) provide a comprehensive overview of BERT’s properties discovered to date.

While recent results indicate that BERT successfully represents lexical-semantic and grammatical information, the evidence of its high-level semantic capabilities is inconclusive. Tenney et al. (2019) show that the English PropBank semantics can be extracted from the encoder and follows syntax in the layer structure. However, out of all formalisms PropBank is most closely tied to syntax, and the results on proto-role and relation probing do not follow the same pattern. Kovaleva et al. (2019) identify two attention heads in BERT responsible for FrameNet relations. However, they find that disabling them in a fine-tuning evaluation on the GLUE Wang et al. (2018) benchmark does not result in decreased performance.

Although we are not aware of any systematic studies dedicated to the effect of formalism on probing results, the evidence of such effects is scattered across the related work: for example, the aforementioned results in Tenney et al. (2019) show a difference in layer utilization between constituents- and dependency-based syntactic probes and semantic role and proto-role probes. It is not clear whether this effect is due to the differences in the underlying datasets and task architecture, or the formalism per se.

Our probing methodology builds upon the edge and layer probing framework. The encoding produced by a frozen BERT model can be seen as a layer-wise snapshot that reflects how the model has constructed the high-level abstractions. Tenney et al. (2019)

introduce the edge probing task design: a simple classifier is tasked with predicting a linguistic property given a pair of spans encoded using a frozen pre-trained model.

Tenney et al. (2019) uses edge probing to analyze the layer utilization of a pre-trained BERT model via scalar mixing weights Peters et al. (2018) learned during training. We revisit this framework in Section 3.

2.3 Role Semantics

We now turn to the object of our investigation: role semantics. For further discussion, consider the following synthetic example:

  1. [label=.]

  2. [John]Ag gave [Mary]Rc a [book]Th.

  3. [Mary]Rc was given a [book]Th by [John]Ag.

Despite surface-level differences, the sentences express the same meaning, suggesting an underlying semantic representation in which these sentences are equivalent. One such representation is offered by role semantics - a shallow predicate-semantic formalism closely related to syntax. In terms of role semantics, Mary, book and John are semantic arguments of the predicate give, and are assigned roles from a pre-defined inventory, for example, Agent, Recipient and Theme.

Semantic roles and their properties have received extensive attention in linguistics Fillmore (1968); Levin and Rappaport Hovav (2005); Dowty (1991) and are considered a universal feature of human language. The size and organization of the role and predicate inventory are subject to debate, giving rise to a variety of role-semantic formalisms.

PropBank assumes a predicate-independent labeling scheme where predicates are distinguished by their sense (get.01), and semantic arguments are labeled with generic numbered core (Arg0-5) and modifier (e.g. AM-TMP) roles. Core roles are not tied to specific definitions, but the effort has been made to keep the role assignments consistent for similar verbs; Arg0 and Arg1 correspond to the Proto-Agent and Proto-Patient roles as per Dowty (1991). The semantic interpretation of core roles depends on the predicate sense.

VerbNet follows a different categorization scheme. Motivated by the regularities in verb behavior, Levin (1993) has introduced the grouping of verbs into intersective classes (ILC). This methodology has been adopted by VerbNet: for example, the VerbNet class get-13.5.1 would include verbs earn, fetch, gain etc. A verb in VerbNet can belong to several classes corresponding to different senses; each class is associated with a set of roles and licensed syntactic transformations. Unlike PropBank, VerbNet uses a set of approx. 30 thematic roles that have universal definitions and are shared among predicates, e.g. Agent, Beneficiary, Instrument.

FrameNet takes a meaning-driven stance on the role encoding by modeling it in terms of frame semantics: predicates are grouped into frames (e.g. Commerce_buy), which specify role-like slots to be filled. FrameNet offers fine-grained frame distinctions, and roles in FrameNet are frame-specific, e.g. Buyer, Seller and Money. The resource accompanies each frame with a description of the situation and its core and peripheral participants.

SPR follows the work of Dowty (1991) and discards the notion of categorical semantic roles in favor of feature bundles. Instead of a fixed role label, each argument is assessed via a 11-dimensional cardinal feature set including Proto-Agent and Proto-Patient properties like volitional, sentient, destroyed, etc. The feature-based approach eliminates some of the theoretical issues associated with categorical role inventories and allows for more flexible modeling of role semantics.

Each of the role labeling formalisms offers certain advantages and disadvantages Giuglea and Moschitti (2006); Mújdricza-Maydt et al. (23-28). While being close to syntax and thereby easier to predict, PropBank doesn’t contribute much semantics to the representation. On the opposite side of the spectrum, FrameNet offers rich predicate-semantic representations for verbs and nouns, but suffers from high granularity and coverage gaps Hartmann et al. (2017). VerbNet takes a middle ground by following grammatical criteria while still encoding coarse-grained semantics, but only focuses on verbs and core (not modifier) roles. SPR avoids the granularity-generalization trade-off of the categorical inventories, but is yet to find its way into practical NLP applications.

3 Probing Methodology

We take the edge probing setup by Tenney et al. (2019) as our starting point. Edge probing aims to predict a label given a pair of contextualized span or word encodings. More formally, we encode a WP-tokenized sentence with a frozen pre-trained model, producing contextual embeddings , each of which is a layered representation over layers, with encoding at layer for the wordpiece further denoted as . A trainable scalar mix is applied to the layered representation to produce the final encoding given the per-layer mixing weights and a scaling parameter :

Given the source and target wordpieces encoded as and , our goal is to predict the label .

Due to its task-agnostic architecture, edge probing can be applied to a wide variety of unary (by omitting ) and binary labeling tasks in a unified manner, facilitating the cross-task comparison. The original setup has several limitations that we address in our implementation.

Regression tasks.

The original edge probing only considers classification tasks. Many language phenomena - including positional information and semantic proto-roles, are naturally modeled as regression, and we extend the original model by supporting both classification and regression: the former achieved via softmax, the latter via direct linear regression to the target value.

Flat model To decrease the models’ own expressive power Hewitt and Liang (2019), we keep the number of parameters in our probing model as low as possible. While Tenney et al. (2019) utilize pooled self-attentional span representations and a projection layer to enable cross-model comparison, we directly feed the wordpiece encoding into the classifier, using the first wordpiece of a word. To further increase the selectivity of the model, we directly project the source and target wordpiece representations into the label space, opposed to the two-layer MLP classifier used in the original setup.

Separate scalar mixes. To enable fine-grained analysis of probing results, we train and analyze separate scalar mixes for source and target wordpieces, motivated by the fact that the classifier might utilize different aspects of their representation for prediction111The original work (Tenney et al., 2019, Appendix C), also uses separate projections for source and target tokens in the background, but does not investigate the differences between the learned projections.. Indeed, we find that the mixing weights learned for source and target wordpieces might show substantial – and linguistically meaningful – variation.

Sentence-level probes. Utilizing the BERT-specific sentence representation [CLS] allows us to incorporate the sentence-level natural language inference (NLI) probe into our kit.

Anchor tasks

We employ two analytical tools from the original layer probing setup. Mixing weight plotting compares layer utilization among tasks by visually aligning the respective learned weight distributions transformed via a softmax function. Layer center-of-gravity is used as a summary statistic for a task’s layer utilization.

While the distribution of mixing weights along the layers allows us to estimate the order in which information is processed during encoding, it doesn’t allow to directly assess the

similarity between the layer utilization of the probing tasks.

Tenney et al. (2019) have demonstrated that the order in which linguistic information is stored in BERT mirrors the traditional NLP pipeline. A prominent property of the NLP pipelines is their use of low-level features to predict downstream phenomena. In the context of layer probing, probing tasks can be seen as end-to-end feature extractors. Following this intuition, we define two groups of probing tasks: target tasks – the main tasks under investigation, and anchor tasks

– a set of related tasks that serve as a basis for qualitative comparison between the targets. The softmax transformation of the scalar mixing weights allows to treat them as probability distributions: the higher the mixing weight of a layer, the more likely the probe is to utilize information from this layer during prediction. We use Kullback-Leibler divergence to compare target tasks (e.g. role labeling in different formalisms) in terms of their similarity to lower-level anchor tasks (e.g. dependency relation and lemma). Note that the notion of anchor task is contextual: the same task can serve as a target and as an anchor, depending on the focus of the study.

4 Setup

4.1 Source data

For German we use the SR3de corpus Mújdricza-Maydt et al. (23-28) that contains parallel PropBank, FrameNet and VerbNet annotations for verbal predicates. For English, SemLink Bonial et al. (2013) provides mappings from the original PropBank corpus annotations to the corresponding FrameNet and VerbNet senses and semantic roles. We use these mappings to enrich the CoNLL-2009 Hajič et al. (2009) dependency role labeling data – also based on the original PropBank – with roles in all three formalisms via a semi-automatic token alignment procedure. The resulting corpus is substantially smaller than the original, but still an order of magnitude larger than SR3de (Table 1). Both corpora are richly annotated with linguistic phenomena on word level, including part-of-speech, lemma and syntactic dependencies. The XNLI probe is sourced from the corresponding development split of the XNLI Conneau et al. (2018) dataset. The SPR probing tasks are extracted from the original data by Reisinger et al. (2015).

tok sent pred arg
CoNLL+SL 312.2K 11.3K 13.3K 23.9K
SR3de 62.6K 2.8K 2.9K 5.5K
Table 1: Statistics for CoNLL+SemLink (English) and SR3de (German), only core roles.
type en de
*token.ix unary 208.9K 46.9K
ttype [v] unary 177.2K 34.0K
lex.unit [v] unary 187.6K 35.7K
pos unary 312.2K 62.6K
deprel binary 300.9K 59.8K
role binary 23.9K 5.5K
**spr binary 9.7K -
xnli unary 2.5K 2.5K
Table 2: Probing task statistics. Tasks marked with [v] use a most frequent label vocabulary. Here and further, tasks marked with * are regression tasks.
language en de
PropBank 5 10
VerbNet 23 29
FrameNet 189 300
Table 3: of role probe labels by formalism.

4.2 Probing tasks

Our probing kit spans a wide range of probing tasks, ranging from primitive surface-level tasks mostly utilized as anchors later to high-level semantic tasks that aim to provide a representational upper bound to predicate semantics. We follow the training, test and development splits from the original SR3de, CoNLL-2009 and SPR data. The XNLI task is sourced from the development set and only used for scalar mix analysis. To reduce the number of labels in some of the probing tasks, we collect frequency statistics over the corresponding training sets and only consider up to 250 most frequent labels. Below we define the tasks in order of their complexity, Table 2 provides the probing task statistics, Table 3 compares the categorical role labeling formalisms in terms of granularity, and Table 4 provides examples. We evaluate the classification performance using Accuracy, while regression tasks are scored via Mean Squared Error.

task input label
token.ix I [saw] a cat. 2
ttype I [saw] a cat. saw
lex.unit I [saw] a cat. see.V
pos I [saw] a cat. VBD
deprel [I]tgt [saw]src a cat. SBJ [I]tgt [saw]src a cat. Experiencer
spr.vltn [I]tgt [saw]src a cat. 2
Table 4: Word-level probing task examples for English. vltn corresponds to the volition SPR property.

Token type (ttype)

predicts the type of a word. This requires contextual processing since a word might consist of several wordpieces;

Token position (token.ix) predicts the linear position of a word, cast as a regression task over the first 20 words in the sentence. Again, the task is non-trivial since it requires the words to be assembled from the wordpieces.

Part-of-speech (pos) predicts the language-specific part-of-speech tag for the given token.

Lexical unit (lex.unit) predicts the lemma and POS of the given word – a common input representation for the entries in lexical resources. We extract coarse POS tags by using the first character of the language-specific POS tag.

Dependency relation (deprel) predicts the dependency relation between the parent src and dependent tgt tokens;

Semantic role (role.[frm]) predicts the semantic role given a predicate src and an argument tgt token in one of the three role labeling formalisms: PropBank pb, VerbNet vn and FrameNet fn. Note that we only probe for the role label, and the model has no access to the verb sense information from the data.

Semantic proto-role (spr.[prop]) is a set of eleven regression tasks predicting the values of the proto-role properties as defined in Reisinger et al. (2015), given a predicate src and an argument tgt.

XNLI is a sentence-level NLI task directly sourced from the corresponding dataset. Given two sentences, the goal is to determine whether an entailment or a contradiction relationship holds between them. We use NLI to investigate the layer utilization of mBERT for high-level semantic tasks. We extract the sentence pair representation via the [CLS] token and treat it as a unary probing task.

5 Results

Our models are implemented using AllenNLP.222We make the code publicly available

We train the probes for 20 epochs using the Adam optimizer with default parameters and a batch size of 32. Due to the frozen encoder and flat model architecture, the total runtime of the experiments is under 8 hours on a single Tesla V100 GPU.

5.1 General Trends

task en de task en de
*token.ix 1.71 2.48 deprel 0.95 0.95
ttype 1.00 1.00 role.fn 0.92 0.59
lex.unit 1.00 1.00 role.pb 0.96 0.71
pos 0.97 0.97 0.94 0.73
Table 5: Best dev score for word-level tasks over 20 epochs ( for classification, for regression).

While absolute performance is secondary to our analysis, we report the probing task scores on respective development sets in Table 5. We observe that grammatical tasks score high, while core role labeling lags behind - in line with the findings of Tenney et al. (2019)333Our results are not directly comparable due to the differences in datasets and formalisms. We observe lower scores for German role labeling which we attribute to the lack of training data. Surprisingly, as we show below, this doesn’t prevent the edge probe from learning to locate relevant role-semantic information in mBERT’s layers.

Our results mirror the findings of Tenney et al. (2019) about the sequential processing order in BERT. We observe that the layer utilization among tasks (Fig. 2) aligns for English and German444Echoing the recent findings on mBERT’s multilingual capacity Pires et al. (2019); Kondratyuk and Straka (2019), although we note that in terms of center-of-gravity mBERT tends to utilize deeper layers for German probes. Basic word-level tasks are indeed processed early by the model, and XNLI probes focus on deeper levels, suggesting that the representation of higher-level semantic phenomena follows the encoding of syntax and predicate semantics.

5.2 The Effect of Formalism

Figure 2: Layer probing results

Using separate scalar mixes for source and target tokens allows us to explore the cross-formalism encoding of role semantics by mBERT in detail. Role labeling probe’s layer utilization drastically differs for predicate and argument tokens. While the argument representation role*tgt mostly focuses on the same layers as the dependency parsing probe, the layer utilization of the predicates role*src is affected by the chosen formalism. PropBank predicate token mixing weights emphasize the same layers as dependency parsing – in line with the previously published results. However, the probes for VerbNet and FrameNet predicates ( src and role.fn src) utilize the layers associated with ttype and lex.unit that contain lexical information. Coupled with the fact that both VerbNet and FrameNet assign semantic roles based on lexical-semantic predicate groupings (frames in FrameNet and verb classes in VerbNet), this suggests that the lower layers of mBERT implicitly encode predicate sense information; moreover, sense encoding for VerbNet utilizes deeper layers of the model associated with syntax, in line with VerbNet’s predicate classification strategy. This finding confirms that the formalism can indeed have linguistically meaningful effects on probing results.

5.3 Anchor Tasks in the Pipeline

Figure 3: Anchor task analysis of SRL formalisms.

We now use the scalar mixes of the role labeling probes as target tasks, and lower-level probes as anchor tasks to qualitatively explore the differences between how our role probes learn to represent predicates and semantic arguments555Darker color corresponds to higher similarity. (Fig. 3). The results reveal a distinctive pattern: while the predicate layer utilization src is similar to the scalar mixes learned for ttype and lex.unit, the learned argument representations tgt attend to the layers associated with dependency relation and POS probes, and the pattern reproduces for English and German. This aligns with the traditional separation of the semantic role labeling task into predicate disambiguation followed by semantic argument identification and labeling, along with the feature sets employed for these tasks Björkelund et al. (2009). Note that the observation about the pipeline-like task processing within the BERT encoders thereby holds, albeit on a sub-task level.

5.4 Formalism Implementations

Both layer and anchor task analysis reveal a prominent discrepancy between English and German role probing results: while the PropBank predicate layer utilization for English mostly relies on syntactic information, German PropBank predicates behave similarly to VerbNet and FrameNet. The difference in the number of role labels for English and German PropBank (Table 3) points at possible qualitative differences in the labeling schemes. The data for English stems from the token-level alignment in SemLink that maps the original PropBank roles to VerbNet and FrameNet. Role annotations for German have a different lineage: they originate from the FrameNet-annotated SALSA corpus Burchardt et al. (2006) semi-automatically converted to PropBank style for the CoNLL-2009 shared task Hajič et al. (2009), and enriched with VerbNet labels in SR3de Mújdricza-Maydt et al. (23-28). As a result, while English PropBank labels are assigned in a predicate-independent manner, German PropBank, following the same numbered labeling scheme, keeps this scheme consistent within the frame. We assume that this incentivizes the probe to learn semantic verb groupings and reflects in our probing results. The ability of the probe to detect subtle differences between formalism implementations constitutes a new use case for probing, and a promising direction for future studies.

5.5 Encoding of Proto-Roles

We now turn to the probing results for decompositional semantic proto-role labeling tasks. Unlike Tenney et al. (2019) who used a multi-label classification probe, we treat SPR properties as separate regression tasks. The results in Table 6 show that the performance varies by property, with some of the properties attaining reasonably low MSE scores despite the simplicity of the probe architecture and the small dataset size. We do not observe a clear performance trend depending on whether the property is associated with Proto-Agent or Patient.

property MSE
(A) *instigation 1.12
(A) *volition 0.87
(A) *awareness 0.79
(A) *sentient 0.47
(A) *change.of.location 0.66
(A) * 1.29
(P) *created 0.87
(P) *destroyed 0.61
(P) *changes.possession 0.83
(P) *change.of.state 1.24
(P) *stationary 0.77
Table 6: Best dev MSE for proto-role probing tasks over 20 epochs. A - Proto-Agent, P - Proto-Patient.

Our fine-grained, property-level task design allows for more detailed insights into the layer utilization by the SPR probes (Fig. 4). The results indicate that while the layer utilization on the predicate side (src) shows no clear preference for particular layers (similar to the results obtained by Tenney et al. (2019)), some of the proto-role features follow the pattern seen in the categorical role labeling and dependency parsing tasks for the argument tokens tgt. With few exceptions, we observe that the properties displaying that behavior are Proto-Agent properties; moreover, a close examination of the results on syntactic preference by Reisinger et al. (2015, p. 483)

reveals that these properties are also the ones with strong preference for the subject position, including the outlier case of

stationery which in their data behaves like a Proto-Agent property. The correspondence is not strict, and we leave an in-depth investigation of the reasons behind these discrepancies for follow-up work.

Figure 4: Layer utilization for SPR properties.

6 Conclusion

We have demonstrated that the choice of linguistic formalism can have substantial, linguistically meaningful effects on role-semantic probing results. We have shown how probing classifiers can be used to detect discrepancies between formalism implementations, and presented evidence of semantic proto-role encoding in the pre-trained mBERT model. Our refined implementation of the edge probing framework coupled with the anchor task methodology enabled new insights into the processing of predicate-semantic information within mBERT. Our findings show that linguistic formalism is an important factor to be accounted for in probing studies. While our work illustrates this point using a single task and a single probing framework, the influence of linguistic formalism per se is likely to be present for any probing setup that builds upon linguistic material. An investigation of how, whether, and why formalisms affect probing results for tasks beyond role labeling and for frameworks beyond edge probing constitutes an exciting avenue for future research.


  • C. F. Baker, C. J. Fillmore, and J. B. Lowe (1998) The Berkeley FrameNet project. In 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics, Volume 1, Montreal, Quebec, Canada, pp. 86–90. External Links: Link, Document Cited by: §1.
  • Y. Belinkov, N. Durrani, F. Dalvi, H. Sajjad, and J. Glass (2017)

    What do neural machine translation models learn about morphology?

    ACL 2017 - 55th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers) 1, pp. 861–872. External Links: Document, 1704.03471, ISBN 9781945626753, Link Cited by: §2.2.
  • A. Björkelund, L. Hafdell, and P. Nugues (2009) Multilingual semantic role labeling. In Proceedings of the Thirteenth Conference on Computational Natural Language Learning (CoNLL 2009): Shared Task, Boulder, Colorado, pp. 43–48. External Links: Link Cited by: §5.3.
  • C. Bonial, K. Stowe, and M. Palmer (2013) Renewing and revising SemLink. In

    Proceedings of the 2nd Workshop on Linked Data in Linguistics (LDL-2013): Representing and linking lexicons, terminologies and other language data

    pp. 9–17. External Links: Link Cited by: §1, §4.1.
  • A. Burchardt, K. Erk, A. Frank, A. Kowalski, S. Padó, and M. Pinkal (2006)

    The SALSA corpus: a German corpus resource for lexical semantics. In Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC’06), Genoa, Italy. External Links: Link Cited by: §5.4.
  • A. Conneau, G. Kruszewski, G. Lample, L. Barrault, and M. Baroni (2018)

    What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties

    ACL 2018 - 56th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers) 1, pp. 2126–2136. External Links: Document, 1805.01070, ISBN 9781948087322, Link Cited by: §2.2.
  • A. Conneau, R. Rinott, G. Lample, A. Williams, S. Bowman, H. Schwenk, and V. Stoyanov (2018) XNLI: evaluating cross-lingual sentence representations. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, pp. 2475–2485. External Links: Link, Document Cited by: §4.1.
  • M. de Marneffe and C. D. Manning (2008) The Stanford typed dependencies representation. In Coling 2008: Proceedings of the workshop on Cross-Framework and Cross-Domain Parser Evaluation, Manchester, UK, pp. 1–8. External Links: Link Cited by: §1.
  • J. Devlin, M. Chang, K. Lee, and K. Toutanova (2019) BERT: pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 4171–4186. External Links: Link, Document Cited by: A Matter of Framing: The Impact of Linguistic Formalism on Probing Results, §1, §2.1.
  • D. Dowty (1991) Thematic proto-roles and argument selection. Language 76 (3), pp. 474–496. External Links: Document, ISBN 0160487749, ISSN 00978507 Cited by: §2.3, §2.3, §2.3.
  • C. J. Fillmore (1968) The Case for Case. In Universals in Linguistic Theory, E. Bach and R. T. Harms (Eds.), pp. 1–88. Cited by: §2.3.
  • M. Gardner, J. Grus, M. Neumann, O. Tafjord, P. Dasigi, N. F. Liu, M. Peters, M. Schmitz, and L. Zettlemoyer (2018) AllenNLP: a deep semantic natural language processing platform. In

    Proceedings of Workshop for NLP Open Source Software (NLP-OSS)

    Melbourne, Australia, pp. 1–6. External Links: Link, Document Cited by: §1.
  • A. Giuglea and A. Moschitti (2006) Semantic role labeling via FrameNet, VerbNet and PropBank. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Sydney, Australia, pp. 929–936. External Links: Link, Document Cited by: §2.3.
  • J. Hajič, M. Ciaramita, R. Johansson, D. Kawahara, M. A. Martí, L. Màrquez, A. Meyers, J. Nivre, S. Padó, J. Štěpánek, P. Straňák, M. Surdeanu, N. Xue, and Y. Zhang (2009) The CoNLL-2009 shared task: syntactic and semantic dependencies in multiple languages. In Proceedings of the Thirteenth Conference on Computational Natural Language Learning (CoNLL 2009): Shared Task, Boulder, Colorado, pp. 1–18. External Links: Link Cited by: §1, §4.1, §5.4.
  • S. Hartmann, I. Kuznetsov, T. Martin, and I. Gurevych (2017) Out-of-domain FrameNet semantic role labeling. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers, Valencia, Spain, pp. 471–482. External Links: Link Cited by: §2.3.
  • J. Hewitt and P. Liang (2019) Designing and interpreting probes with control tasks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, pp. 2733–2743. External Links: Link, Document Cited by: §3.
  • J. Hewitt and C. D. Manning (2019) A structural probe for finding syntax in word representations. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 4129–4138. External Links: Link, Document Cited by: §2.2.
  • K. Kipper-Schuler (2005) VerbNet: A broad-coverage, comprehensive verb lexicon. Ph.D. Thesis, University of Pennsylvania. Cited by: §1.
  • D. Kondratyuk and M. Straka (2019) 75 languages, 1 model: parsing universal dependencies universally. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, pp. 2779–2795. External Links: Link, Document Cited by: footnote 4.
  • O. Kovaleva, A. Romanov, A. Rogers, and A. Rumshisky (2019) Revealing the dark secrets of BERT. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, pp. 4365–4374. External Links: Link, Document Cited by: §1, §2.2.
  • Z. Lan, M. Chen, S. Goodman, K. Gimpel, P. Sharma, and R. Soricut (2019)

    ALBERT: a lite bert for self-supervised learning of language representations

    arXiv:1909.11942. External Links: 1909.11942 Cited by: §2.1.
  • B. Levin and M. Rappaport Hovav (2005) Argument realization. Research Surveys in Linguistics, Cambridge University Press. External Links: Document Cited by: §2.3.
  • B. Levin (1993) English verb classes and alternations: a preliminary investigation. University of Chicago Press. External Links: ISBN 9780226475332, LCCN lc92042504, Link Cited by: §2.3.
  • Y. Lin, Y. C. Tan, and R. Frank (2019) Open sesame: getting inside BERT’s linguistic knowledge. In

    Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP

    Florence, Italy, pp. 241–253. External Links: Link, Document Cited by: §2.2.
  • Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. S. Zettlemoyer, and V. Stoyanov (2019) RoBERTa: a robustly optimized bert pretraining approach. arXiv:1907.11692 abs/1907.11692. Cited by: §2.1.
  • É. Mújdricza-Maydt, S. Hartmann, I. Gurevych, and A. Frank (23-28) Combining semantic annotation of word sense & semantic roles: a novel annotation scheme for VerbNet roles on German language data. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), N. C. (. Chair), K. Choukri, T. Declerck, S. Goggi, M. Grobelnik, B. Maegaard, J. Mariani, H. Mazo, A. Moreno, J. Odijk, and S. Piperidis (Eds.), (english). External Links: ISBN 978-2-9517408-9-1 Cited by: §1, §2.3, §4.1, §5.4.
  • M. Palmer, D. Gildea, and P. Kingsbury (2005) The proposition bank: an annotated corpus of semantic roles. Computational Linguistics 31 (1), pp. 71–106. External Links: Link, Document Cited by: §1.
  • M. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, and L. Zettlemoyer (2018) Deep contextualized word representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), New Orleans, Louisiana, pp. 2227–2237. External Links: Link, Document Cited by: §1, §2.2.
  • T. Pires, E. Schlinger, and D. Garrette (2019) How multilingual is multilingual BERT?. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, pp. 4996–5001. External Links: Link, Document Cited by: footnote 4.
  • C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y. Zhou, W. Li, and P. J. Liu (2019)

    Exploring the limits of transfer learning with a unified text-to-text transformer

    arXiv:1910.10683. External Links: 1910.10683 Cited by: §2.1.
  • E. Reif, A. Yuan, M. Wattenberg, F. B. Viegas, A. Coenen, A. Pearce, and B. Kim (2019) Visualizing and measuring the geometry of bert. In Advances in Neural Information Processing Systems 32, H. Wallach, H. Larochelle, A. Beygelzimer, F. d’Alché-Buc, E. Fox, and R. Garnett (Eds.), pp. 8594–8603. External Links: Link Cited by: §1, §2.2.
  • D. Reisinger, R. Rudinger, F. Ferraro, C. Harman, K. Rawlins, and B. Van Durme (2015) Semantic proto-roles. Transactions of the Association for Computational Linguistics 3, pp. 475–488. External Links: Link, Document Cited by: §1, §4.1, §4.2, §5.5.
  • A. Rogers, O. Kovaleva, and A. Rumshisky (2020) A primer in bertology: what we know about how bert works. arXiv:2002.12327. Cited by: §1, §2.2.
  • S. Schuster and C. D. Manning (2016) Enhanced English universal dependencies: an improved representation for natural language understanding tasks. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), Portorož, Slovenia, pp. 2371–2378. External Links: Link Cited by: §1.
  • I. Tenney, D. Das, and E. Pavlick (2019) BERT rediscovers the classical NLP pipeline. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, pp. 4593–4601. External Links: Link, Document Cited by: §2.2, §2.2, §2.2, §2.2, §3, §5.1, §5.1, §5.5.
  • I. Tenney, P. Xia, B. Chen, A. Wang, A. Poliak, R. Thomas McCoy, N. Kim, B. Van Durme, S. R. Bowman, D. Das, and E. Pavlick (2019) What do you learn from context? Probing for sentence structure in contextualized word representations. 7th International Conference on Learning Representations, ICLR 2019, pp. 1–17. External Links: 1905.06316, Link Cited by: §1, §1, §2.2, §3, §3, §5.5, footnote 1.
  • A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin (2017) Attention is all you need. In Advances in Neural Information Processing Systems 30, I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.), pp. 5998–6008. External Links: Link Cited by: §2.1.
  • A. Wang, A. Singh, J. Michael, F. Hill, O. Levy, and S. Bowman (2018) GLUE: a multi-task benchmark and analysis platform for natural language understanding. In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, Brussels, Belgium, pp. 353–355. External Links: Link, Document Cited by: §2.2, §2.2.
  • A. S. White, D. Reisinger, K. Sakaguchi, T. Vieira, S. Zhang, R. Rudinger, K. Rawlins, and B. Van Durme (2016) Universal decompositional semantics on universal dependencies. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, Austin, Texas, pp. 1713–1723. External Links: Link, Document Cited by: §1.
  • T. Wolf, L. Debut, V. Sanh, J. Chaumond, C. Delangue, A. Moi, P. Cistac, T. Rault, R. Louf, M. Funtowicz, and J. Brew (2019) HuggingFace’s Transformers: state-of-the-art natural language processing. arXiv:1910.03771. Cited by: §1.
  • Y. Wu, M. Schuster, Z. Chen, Q. V. Le, M. Norouzi, W. Macherey, M. Krikun, Y. Cao, Q. Gao, K. Macherey, J. Klingner, A. Shah, M. Johnson, X. Liu, Ł. Kaiser, S. Gouws, Y. Kato, T. Kudo, H. Kazawa, K. Stevens, G. Kurian, N. Patil, W. Wang, C. Young, J. Smith, J. Riesa, A. Rudnick, O. Vinyals, G. Corrado, M. Hughes, and J. Dean (2016) Google’s neural machine translation system: bridging the gap between human and machine translation. arXiv:1609.08144 abs/1609.08144. External Links: Link Cited by: §2.1.