Annotation Guidelines

How causal relations are annotated in the Causal Semantics framework

Overview

This page presents the annotation schema and guidelines used to produce the training data for C-BERT and the underlying (C, E, I) tuple representation.

If you are using C-BERT or working with annotated data from this project, this page serves as the reference for how annotations are structured, what decisions were made, and how edge cases are handled.

The annotation was developed over three iterations (2024–2025) and primarily conducted via INCEpTION. A total of 4,753 sentences from German environmental discourse (1990–2022) yielded 2,391 manually annotated causal relations.

Annotation Schema

The schema consists of two layers: span annotations mark causal components in text, and relation annotations link them into structured causal relations.

Spans

Each token or token sequence in a sentence receives at most one span label. There are two span categories.

Role marks the primary causal components:

Indicator — the lexical marker that projects the causal relation
- e.g. verursachen ‘cause’, ist Ursache ‘is cause’, weil ‘because’, stoppen ‘stop’
Entity — a cause or effect entity involved in the relation
- e.g. Klimawandel ‘climate change’, Pestizide ‘pesticides’

Coefficient captures semantic modifiers:

Negation
- Neutralizes or inverts the causal relation
- nicht ‘not’, kein ‘no’, Verlust ‘loss’, Schwund ‘shrinkage’
Division
- Lowers salience
- auch ‘also’, unter anderem ‘among others’
Priority
- Increases salience
- vor allem ‘especially’, wichtig ‘important’

The three types above directly impact influence. Other coefficients run orthogonal, as they either form constraints (e.g. temporality) – or specify epistemic dimensions (e.g Uncertainty).

Orthogonal coefficients

Object-based constraints
Coefficient	Function	Examples
`Temporality`	Temporal framing	seit, bereits, künftig, Jahre
`Spatiality`	Spatial framing	global, lokal, weltweit, Deutschland
`Quality`	Qualitative modification	industrielle, parasitischer, schwefelhaltiger
`Quantity`	Quantitative modification	großes, fünftausend, zwölf

Propositional-based constraints
Coefficient	Function	Examples
`Uncertainty`	Modality and hedging	möglicherweise, könnte, vermutlich
`Representation`	Evidentiality marker	laut, sagt, sei
`Representation Entity`	Source/provenance	Studie, Bericht, Forscher

Relations

Annotated spans are linked by directed relations:

Cause: Indicator → Entity (the entity is designated as a Cause)
Effect: Indicator → Entity (the entity fills the Effect role)
Constraint: Entity or Indicator → Coefficient (the coefficient modifies its governor)

Example

Consider the spans in this sentence:

Klimawandel und insbesondere Landwirtschaft verstärken möglicherweise das weltweite Artensterben.

Climate change and especially agriculture possibly intensify global species extinction.

Two coordinated entities (Klimawandel und Landwirtschaft) are annotated separately with parallel Cause relations.

The adverb insbesondere ‘especially’ is scoped to Landwirtschaft, raising its salience as Priority. The sentence adverb möglicherweise (Uncertainty) is parented to the indicator since it modifies the causal proposition.

Lastly, the adjective weltweit is separately encoded as Spatiality coefficient – in accordance with the token minimization principle.

graph LR
    A["*verstärken*"] -->|Cause| B["*Klimawandel*"]
    A -->|Cause| C["*Landwirtschaft*"] -->|Priority| D["*insbesondere*"]
    A -->|Uncertainty| E["*möglicherweise*"]
    A -->|Effect| F["*Artensterben*"]
    F -->|Spatiality| G["*weltweit*"]

Annotation Principles

Four principles guide annotation decisions in ambiguous cases.

Minimal Principle

Only explicitly marked causal relations are annotated — no inference. When a lexically specific marker (e.g. verursacht) and a functional marker (e.g. durch) co-occur, only the lexically richer element is annotated as indicator. Functional prepositions and connectors are annotated as indicators only when no richer causal lexeme is present.

For light verb constructions, only the semantically loaded element is annotated (e.g. hat etwas mit X zu tun → Indicator: tun).

Token Minimization

Entities are reduced to their head token. Attributive modifiers are extracted as separate coefficients:

industrielle Landwirtschaft → Entity: Landwirtschaft, Coefficient: industrielle (Quality)
Einsatz von Pestiziden → Entity: Pestiziden
Exception: named entities (Europäische Union) and fixed multi-word expressions (saurer Regen) are annotated as single units.

This prevents proliferation of marginal entity variants and facilitates downstream aggregation.

Syntactic Proximity

When multiple potential entities compete for a role, the syntactically closest entity to the indicator is preferred, unless semantic considerations override this.

Coefficient Conservatism

Coefficients (other than Representation) are only annotated when they stand in a direct syntactic dependency relation with an indicator or entity.

Indicators

Indicators are the lexical or syntactic markers that project causal relations and establish (C, E, I) tuples. They fulfill two functions: projecting causal roles onto syntactic positions in their co-text, and encoding inherent information about polarity and salience.

The annotation corpus contains 642 distinct indicator forms, grouped into 192 indicator families by morphological and semantic criteria. A family subsumes all realizations of a shared lexical core. E.g. the Ursache family includes, among other:

verursachen ‘cause’
ist Ursache ‘is the cause’,
Ursache sein ‘to be the cause’
Teilursache sein ‘to be a partial cause’

Top 10 Indicator Families

Family	Forms	Instances
Ursache ‘Cause’	21	167
Verantwortung ‘Responsibility’	11	122
Stoppen ‘Stop’	3	111
Gegen ‘Against’	3	95
Durch ‘through’	2	70

Family	Forms	Instances
Beitrag ‘Contribution’	10	70
Folge ‘Consequence’	6	69
Kampf ‘Fight’	8	69
Führen ‘Lead’	9	59
Grund ‘Reason’	7	57

Orientation

Some families are Cause-oriented, others are Effect-oriented. Compare:

Emission_C is a cause of pollution_E.
Pollution_E is a consequence of emission_C.

Orientation affects projection – Cause-oriented indicators generally project Cause on the subject, Effect-oriented indicators do the same with Effect. Modified indicator forms affect the projection according to their orientation.

Note the difference between Teilursache ‘partial cause’ and Teilwirkung partial effect: The former implies the existence of several causes – the latter the existence of several effects.

Polarity and Salience

Each indicator family carries a default polarity (promoting + or inhibiting -) and a default salience class:

Table Caption
Family	Polarity	Default Salience	Discourse Function
Ursache ‘Cause’	+	[0.5-1]	Cause-oriented prototype
Folge ‘Consequence’	+	1	Effect-oriented
Beitrag ‘Contribution’	+	[0.5-0.75]	Distributional attribution
Durch ‘Through’	+	1	Grammaticalized preposition
Stoppen ‘Stop’	−	1	Intervention framing
Gegen ‘Against’	−	1	Variable condensation (Kampf/Protest gegen ‘fight/protest against’)
Reduzieren ‘Reduce’	−	[0.5-0.75]	Prototypical distributional negative
Wirkung ‘Effect’	\pm	[0-1]	Widest range of influence (Gegenwirkung ‘counter-effect’, wirkungslos ‘ineffective’ )

Syntactic Projection Patterns

The syntactic realization of each indicator determines how Cause and Effect roles are projected:

Family	Example	Cause Projection	Effect Projection
Ursache ‘Cause’	X ist Ursache für Y ‘x is the cause of’	Subject	PP (für) ‘for’
Folge ‘Consequence’	Y ist Folge von X ‘x is a consequence of’	PP (von) ‘of’	Subject
Verursachen ’Cause	X verursacht Y ‘x causes y’	Subject	Accusative object
Beitragen ‘Contribution’	X trägt zu Y bei ‘x contributes to y’	Subject	PP (zu) ‘to’
Stoppen ‘Stop’	X stoppt Y ‘x stops y’	Subject	Accusative object
Durch ‘Through’	Y durch X ‘x through y’	PP complement	Head noun
Gegen ‘Against’	X gegen Y ‘x against y’	Subject	PP complement

Context Markers

Context markers modify and contextualize the causal relations projected by indicators. Unlike indicators, they are not lexically specialized for causality but operate at the predication or proposition level. Three structural marker types directly affect the INFLUENCE computation:

Division

Division markers signal polycausal structures with implicit co-causes. Typical realizations include unter anderem ‘among other things’, auch ‘also’, ebenfalls ‘likewise’, and the composite nicht nur ‘not only’.

Division markers reduce the salience to |I| = 0.5 regardless of the indicator’s default.

Priority

Priority markers establish asymmetric weighting within polycausal sets: vor allem ‘above all’, hauptsächlich ‘mainly, maßgeblich ’significantly’. They set |I| = 0.75 for the prioritized cause.

Both marker types affect only the salience (|I|), leaving polarity (\pm) unchanged.

Negation

Negation is the structurally most influential marker. Two types are distinguished:

Object-based negation operates at the entity level through markers like Verlust ‘loss’, Schwund ‘decline’, Rückgang ‘decrease’. These invert the polarity when the count of negations is odd:

Verlust von Lebensräumen verursacht Bienensterben.
Loss of habitat causes bee decline.

Indicator verursachen ‘cause’: default I = 1;
object negation on Cause (Verlust ‘loss’) inverts polarity: I = -1.

Propositional negation operates at the relation level (nicht, kein) and neutralizes the entire causal relationship (I = 0):

Pestizide verursachen nicht Insektensterben.
Pesticides don’t cause Insektensterben.

Indicator verursachen ‘cause’: default I = 1
Propositional negation nicht ‘don’t’ neutralizes: I = 0.

From Annotations to INFLUENCE

The annotations documented above — indicators, entities, and context markers — are the inputs to a deterministic computation that produces the final INFLUENCE value I \in [-1, +1].

In brief: entity identification follows the indicator’s syntactic projection pattern, polarity is determined by indicator class and negation markers, and salience is computed through a cascading hierarchy of morphological, determiner, and syntactic markers.

For the full formal specification — including the cascade rules, coordination normalization, and worked examples — see Tuple Construction.

Data Format

Annotated data is exported as JSON. Each entry represents a sentence with its metadata and extracted relations.

Schema

{
  "subfolder": "Artensterben_oa",
  "global_sentence_id": 1148482,
  "text_id": "FAZ_200204_384209",
  "text_date": "2002-04",
  "sentence_id": "12",
  "sentence": "...",
  "relations": [
    {
      "indicator": "Folge",
      "entities": [
        {
          "entity": "Kleinplanet",
          "relation": "Cause",
          "dependent_coefficients": [
            {"coefficient_text": "Jahren", "coefficient": "Temporality"},
            {"coefficient_text": "Mexikos", "coefficient": "Spatiality"}
          ]
        },
        {
          "entity": "Artensterben",
          "relation": "Effect"
        }
      ],
      "coefficient": "Division",
      "dependent_coefficients": [
        {"coefficient_text": "könnte", "coefficient": "Uncertainty"}
      ],
      "representation": "berieten",
      "representation_entities": [
        {
          "entity": "Teilnehmer",
          "relation": "Constraint",
          "dependent_coefficients": [
            {"coefficient_text": "fünftausend", "coefficient": "Quantity"}
          ]
        }
      ]
    }
  ]
}

Field Reference

Sentence-level fields:

Field	Description
`subfolder`	WABI subcorpus and syntactic position (e.g. `Artensterben_oa` = accusative object)
`global_sentence_id`	Unique sentence identifier across the full corpus
`text_id`	Source document identifier (format: `SOURCE_YYYYMM_ID`)
`text_date`	Publication date (YYYY-MM)
`sentence`	Full sentence text
`relations`	Array of causal relations found in this sentence (empty if none)

Relation-level fields:

Field	Description
`indicator`	The causal indicator lexeme
`entities`	Array of entities with their causal role (`Cause` or `Effect`)
`coefficient`	Structural marker on the relation level (`Negation`, `Division`)
`dependent_coefficients`	Contextual coefficients attached to the indicator
`representation`	Evidentiality/speech-act verb (if present)
`representation_entities`	Source entities for reported speech

Entity-level fields:

Field	Description
`entity`	Head token of the entity (token-minimized)
`relation`	Causal role: `Cause`, `Effect`, or `Constraint`
`dependent_coefficients`	Array of coefficients modifying this entity

Sentences Without Relations

Sentences where no explicit causal relation was identified have an empty relations array. These are not noise — they were reviewed during annotation and determined to contain no explicit causal markers per the minimal principle.

Corpus Statistics

Total sentences	4,753
Sentences with ≥1 WABI relation	1,797 (37.8%)
Total WABI-relevant relations	1,867
Distinct indicator forms	642
Indicator families	192
Mean relations per sentence	0.39

Per WABI term:

Term	Sentences	Relations	Rel./Sent.
Waldsterben	1,818	633	0.35
Artensterben	1,854	744	0.40
Bienensterben	536	257	0.48
Insektensterben	545	233	0.43

--- title: "Annotation Guidelines" subtitle: "How causal relations are annotated in the Causal Semantics framework" --- ## Overview This page presents the annotation schema and guidelines used to produce the training data for [C-BERT](c-bert.qmd) and the underlying $(C, E, I)$ tuple representation. If you are using C-BERT or working with annotated data from this project, this page serves as the reference for how annotations are structured, what decisions were made, and how edge cases are handled. The annotation was developed over three iterations (2024–2025) and primarily conducted via [INCEpTION](https://inception-project.github.io/). A total of 4,753 sentences from German environmental discourse (1990–2022) yielded 2,391 manually annotated causal relations. ## Annotation Schema The schema consists of two layers: **span annotations** mark causal components in text, and **relation annotations** link them into structured causal relations. ### Spans Each token or token sequence in a sentence receives at most one span label. There are two span categories. **Role** marks the primary causal components: - `Indicator` — the lexical marker that projects the causal relation - e.g. *verursachen* 'cause', *ist Ursache* 'is cause', *weil* 'because', *stoppen* 'stop' - `Entity` — a cause or effect entity involved in the relation - e.g. *Klimawandel* 'climate change', *Pestizide* 'pesticides' **Coefficient** captures semantic modifiers: - `Negation` - Neutralizes or inverts the causal relation - *nicht* 'not', *kein* 'no', *Verlust* 'loss', *Schwund* 'shrinkage' - `Division` - Lowers salience - *auch* 'also', *unter anderem* 'among others' - `Priority` - Increases salience - *vor allem* 'especially', *wichtig* 'important' The three types above directly impact influence. Other coefficients run orthogonal, as they either form constraints (e.g. `temporality`) – or specify epistemic dimensions (e.g `Uncertainty`). ::: {.callout-note collapse="true"} ## Orthogonal coefficients | Coefficient | Function | Examples | | ----------------------- | ------------------------- | --------------------------------------------------- | | `Temporality` | Temporal framing | *seit*, *bereits*, *künftig*, *Jahre* | | `Spatiality` | Spatial framing | *global*, *lokal*, *weltweit*, *Deutschland* | | `Quality` | Qualitative modification | *industrielle*, *parasitischer*, *schwefelhaltiger* | | `Quantity` | Quantitative modification | *großes*, *fünftausend*, *zwölf* | : Object-based constraints | Coefficient | Function | Examples | | ----------------------- | ------------------------- | --------------------------------------------------- | | `Uncertainty` | Modality and hedging | *möglicherweise*, *könnte*, *vermutlich* | | `Representation` | Evidentiality marker | *laut*, *sagt*, *sei* | | `Representation Entity` | Source/provenance | *Studie*, *Bericht*, *Forscher* | : Propositional-based constraints ::: ### Relations Annotated spans are linked by directed relations: - `Cause`: Indicator → Entity (the entity is designated as a [Cause]{.smallcaps}) - `Effect`: Indicator → Entity (the entity fills the [Effect]{.smallcaps} role) - `Constraint`: Entity or Indicator → Coefficient (the coefficient modifies its governor) ### Example Consider the spans in this sentence: ::: {layout-ncol="2"} :::: {} > [Klimawandel]{.entity} und [insbesondere]{.priority} [Landwirtschaft]{.entity} [verstärken]{.indicator} [möglicherweise]{.uncertainty} das [weltweite]{.spatiality} [Artensterben]{.entity}. > [Climate change]{.entity} and [especially]{.priority} [agriculture]{.entity} [possibly]{.uncertainty} [intensify]{.indicator} [global]{.spatiality} [species extinction]{.entity}. :::: :::: {} Two coordinated entities (*Klimawandel und Landwirtschaft*) are annotated separately with parallel `Cause` relations. The adverb *insbesondere* 'especially' is scoped to *Landwirtschaft*, raising its salience as ``Priority``. The sentence adverb *möglicherweise* (`Uncertainty`) is parented to the indicator since it modifies the causal proposition. Lastly, the adjective *weltweit* is separately encoded as `Spatiality` coefficient – in accordance with the token minimization principle. :::: ::: ```{mermaid} graph LR A["*verstärken*"] -->|Cause| B["*Klimawandel*"] A -->|Cause| C["*Landwirtschaft*"] -->|Priority| D["*insbesondere*"] A -->|Uncertainty| E["*möglicherweise*"] A -->|Effect| F["*Artensterben*"] F -->|Spatiality| G["*weltweit*"] ``` ## Annotation Principles Four principles guide annotation decisions in ambiguous cases. ### Minimal Principle Only explicitly marked causal relations are annotated — no inference. When a lexically specific marker (e.g. *verursacht*) and a functional marker (e.g. *durch*) co-occur, only the lexically richer element is annotated as indicator. Functional prepositions and connectors are annotated as indicators only when no richer causal lexeme is present. For light verb constructions, only the semantically loaded element is annotated (e.g. *hat etwas mit X **zu tun*** → Indicator: *tun*). ### Token Minimization Entities are reduced to their head token. Attributive modifiers are extracted as separate coefficients: - *industrielle Landwirtschaft* → Entity: `Landwirtschaft`, Coefficient: `industrielle` (Quality) - *Einsatz von Pestiziden* → Entity: `Pestiziden` - Exception: named entities (*Europäische Union*) and fixed multi-word expressions (*saurer Regen*) are annotated as single units. This prevents proliferation of marginal entity variants and facilitates downstream aggregation. ### Syntactic Proximity When multiple potential entities compete for a role, the syntactically closest entity to the indicator is preferred, unless semantic considerations override this. ### Coefficient Conservatism Coefficients (other than `Representation`) are only annotated when they stand in a direct syntactic dependency relation with an indicator or entity. ## Indicators Indicators are the lexical or syntactic markers that project causal relations and establish $(C, E, I)$ tuples. They fulfill two functions: projecting causal roles onto syntactic positions in their co-text, and encoding inherent information about polarity and salience. The annotation corpus contains **642 distinct indicator forms**, grouped into **192 indicator families** by morphological and semantic criteria. A family subsumes all realizations of a shared lexical core. E.g. the [Ursache]{.smallcaps} family includes, among other: - *verursachen* 'cause' - *ist Ursache* 'is the cause', - *Ursache sein* 'to be the cause' - *Teilursache sein* 'to be a partial cause' ### Top 10 Indicator Families ::: {layout-ncol=2} | Family | Forms | Instances | |--------|-------|-----------| | [Ursache]{.smallcaps} '[Cause]{.smallcaps}' | 21 | 167 | | [Verantwortung]{.smallcaps} '[Responsibility]{.smallcaps}' | 11 | 122 | | [Stoppen]{.smallcaps} '[Stop]{.smallcaps}' | 3 | 111 | | [Gegen]{.smallcaps} '[Against]{.smallcaps}' | 3 | 95 | | [Durch]{.smallcaps} '[through]{.smallcaps}' | 2 | 70 | | Family | Forms | Instances | |--------|-------|-----------| | [Beitrag]{.smallcaps} '[Contribution]{.smallcaps}' | 10 | 70 | | [Folge]{.smallcaps} '[Consequence]{.smallcaps}' | 6 | 69 | | [Kampf]{.smallcaps} '[Fight]{.smallcaps}' | 8 | 69 | | [Führen]{.smallcaps} '[Lead]{.smallcaps}' | 9 | 59 | | [Grund]{.smallcaps} '[Reason]{.smallcaps}' | 7 | 57 | ::: ::: {.callout-note collapse="true"} ## Orientation Some families are [Cause]{.smallcaps}-oriented, others are [Effect]{.smallcaps}-oriented. Compare: > *Emission$_C$ is a cause of pollution$_E$.* > *Pollution$_E$ is a consequence of emission$_C$.* Orientation affects projection – [Cause]{.smallcaps}-oriented indicators generally project [Cause]{.smallcaps} on the subject, [Effect]{.smallcaps}-oriented indicators do the same with [Effect]{.smallcaps}. Modified indicator forms affect the projection according to their orientation. Note the difference between *Teilursache* 'partial cause' and *Teilwirkung* *partial effect*: The former implies the existence of several causes – the latter the existence of several effects. ::: ### Polarity and Salience Each indicator family carries a default **polarity** (promoting $+$ or inhibiting $-$) and a default **salience** class: | Family | Polarity | Default Salience | Discourse Function | | ------------------------------------------------------- | -------- | ---------------- | -------------------------------------------------------------------------- | | [Ursache]{.smallcaps} '[Cause]{.smallcaps}' | $+$ | $[0.5-1]$ | [Cause]{.smallcaps}-oriented prototype | | [Folge]{.smallcaps} '[Consequence]{.smallcaps}' | $+$ | $1$ | [Effect]{.smallcaps}-oriented | | [Beitrag]{.smallcaps} '[Contribution]{.smallcaps}' | $+$ | $[0.5-0.75]$ | Distributional attribution | | [Durch]{.smallcaps} '[Through]{.smallcaps}' | $+$ | $1$ | Grammaticalized preposition | | [Stoppen]{.smallcaps} '[Stop]{.smallcaps}' | $−$ | $1$ | Intervention framing | | [Gegen]{.smallcaps} '[Against]{.smallcaps}' | $−$ | $1$ | Variable condensation (*Kampf/Protest gegen* 'fight/protest against') | | [Reduzieren]{.smallcaps} '[Reduce]{.smallcaps}' | $−$ | $[0.5-0.75]$ | Prototypical distributional negative | | [Wirkung]{.smallcaps} '[Effect]{.smallcaps}' | $\pm$ | $[0-1]$ | Widest range of influence (*Gegenwirkung* 'counter-effect', *wirkungslos* 'ineffective' ) | : Table Caption {tbl-colwidths="[30, 15, 25, 40]"} ### Syntactic Projection Patterns The syntactic realization of each indicator determines how [Cause]{.smallcaps} and [Effect]{.smallcaps} roles are projected: | Family | Example | Cause Projection | Effect Projection | | --------------------------------------------------------- | ------------------------------------------------ | -------------------- | --------------------- | | [Ursache]{.smallcaps} '[Cause]{.smallcaps}' | *X ist Ursache für Y* 'x is the cause of' | Subject | PP (*für*) 'for' | | [Folge]{.smallcaps} '[Consequence]{.smallcaps}' | *Y ist Folge von X* 'x is a consequence of' | PP (*von*) 'of' | Subject | | [Verursachen]{.smallcaps} '[Cause]{.smallcaps} | *X verursacht Y* 'x causes y' | Subject | Accusative object | | [Beitragen]{.smallcaps} '[Contribution]{.smallcaps}' | *X trägt zu Y bei* 'x contributes to y' | Subject | PP (*zu*) 'to' | | [Stoppen]{.smallcaps} '[Stop]{.smallcaps}' | *X stoppt Y* 'x stops y' | Subject | Accusative object | | [Durch]{.smallcaps} '[Through]{.smallcaps}' | *Y durch X* 'x through y' | PP complement | Head noun | | [Gegen]{.smallcaps} '[Against]{.smallcaps}' | *X gegen Y* 'x against y' | Subject | PP complement | ## Context Markers Context markers modify and contextualize the causal relations projected by indicators. Unlike indicators, they are not lexically specialized for causality but operate at the predication or proposition level. Three structural marker types directly affect the INFLUENCE computation: ### Division Division markers signal polycausal structures with implicit co-causes. Typical realizations include *unter anderem* 'among other things', *auch* 'also', *ebenfalls* 'likewise', and the composite *nicht nur* 'not only'. Division markers reduce the salience to $|I| = 0.5$ regardless of the indicator's default. ### Priority Priority markers establish asymmetric weighting within polycausal sets: *vor allem* 'above all', *hauptsächlich* 'mainly, *maßgeblich* 'significantly'. They set $|I| = 0.75$ for the prioritized cause. Both marker types affect only the salience ($|I|$), leaving polarity ($\pm$) unchanged. ### Negation Negation is the structurally most influential marker. Two types are distinguished: **Object-based negation** operates at the entity level through markers like *Verlust* 'loss', *Schwund* 'decline', *Rückgang* 'decrease'. These invert the polarity when the count of negations is odd: > ***Verlust** von Lebensräumen verursacht Bienensterben.* > ***Loss of** habitat causes bee decline.* > > Indicator *verursachen* 'cause': default $I = 1$; > object negation on [Cause]{.smallcaps} (*Verlust* 'loss') inverts polarity: $I = -1$. **Propositional negation** operates at the relation level (*nicht*, *kein*) and neutralizes the entire causal relationship ($I = 0$): > *Pestizide verursachen **nicht** Insektensterben.* > *Pesticides **don't** cause Insektensterben.* > > Indicator *verursachen* 'cause': default $I = 1$ > Propositional negation *nicht* 'don't' neutralizes: $I = 0$. ## From Annotations to INFLUENCE The annotations documented above — indicators, entities, and context markers — are the inputs to a deterministic computation that produces the final INFLUENCE value $I \in [-1, +1]$. In brief: entity identification follows the indicator's syntactic projection pattern, polarity is determined by indicator class and negation markers, and salience is computed through a cascading hierarchy of morphological, determiner, and syntactic markers. For the full formal specification — including the cascade rules, coordination normalization, and worked examples — see [Tuple Construction](../processing/tuple-construction.qmd). ## Data Format {#data-format} Annotated data is exported as JSON. Each entry represents a sentence with its metadata and extracted relations. ### Schema ```json { "subfolder": "Artensterben_oa", "global_sentence_id": 1148482, "text_id": "FAZ_200204_384209", "text_date": "2002-04", "sentence_id": "12", "sentence": "...", "relations": [ { "indicator": "Folge", "entities": [ { "entity": "Kleinplanet", "relation": "Cause", "dependent_coefficients": [ {"coefficient_text": "Jahren", "coefficient": "Temporality"}, {"coefficient_text": "Mexikos", "coefficient": "Spatiality"} ] }, { "entity": "Artensterben", "relation": "Effect" } ], "coefficient": "Division", "dependent_coefficients": [ {"coefficient_text": "könnte", "coefficient": "Uncertainty"} ], "representation": "berieten", "representation_entities": [ { "entity": "Teilnehmer", "relation": "Constraint", "dependent_coefficients": [ {"coefficient_text": "fünftausend", "coefficient": "Quantity"} ] } ] } ] } ``` ### Field Reference **Sentence-level fields:** | Field | Description | |---|---| | `subfolder` | WABI subcorpus and syntactic position (e.g. `Artensterben_oa` = accusative object) | | `global_sentence_id` | Unique sentence identifier across the full corpus | | `text_id` | Source document identifier (format: `SOURCE_YYYYMM_ID`) | | `text_date` | Publication date (YYYY-MM) | | `sentence` | Full sentence text | | `relations` | Array of causal relations found in this sentence (empty if none) | **Relation-level fields:** | Field | Description | |---|---| | `indicator` | The causal indicator lexeme | | `entities` | Array of entities with their causal role (`Cause` or `Effect`) | | `coefficient` | Structural marker on the relation level (`Negation`, `Division`) | | `dependent_coefficients` | Contextual coefficients attached to the indicator | | `representation` | Evidentiality/speech-act verb (if present) | | `representation_entities` | Source entities for reported speech | **Entity-level fields:** | Field | Description | |---|---| | `entity` | Head token of the entity (token-minimized) | | `relation` | Causal role: `Cause`, `Effect`, or `Constraint` | | `dependent_coefficients` | Array of coefficients modifying this entity | ### Sentences Without Relations Sentences where no explicit causal relation was identified have an empty `relations` array. These are not noise — they were reviewed during annotation and determined to contain no explicit causal markers per the minimal principle. ## Corpus Statistics | | | |---|---| | **Total sentences** | 4,753 | | **Sentences with ≥1 WABI relation** | 1,797 (37.8%) | | **Total WABI-relevant relations** | 1,867 | | **Distinct indicator forms** | 642 | | **Indicator families** | 192 | | **Mean relations per sentence** | 0.39 | **Per WABI term:** | Term | Sentences | Relations | Rel./Sent. | |---|---|---|---| | Waldsterben | 1,818 | 633 | 0.35 | | Artensterben | 1,854 | 744 | 0.40 | | Bienensterben | 536 | 257 | 0.48 | | Insektensterben | 545 | 233 | 0.43 | ## Further Reading  - For how annotations are transformed into $(C, E, I)$ tuples, see [Tuple Construction](../processing/tuple-construction.qmd) - For the C-BERT model trained on this data, see [C-BERT](c-bert.qmd) - Full annotation data (Bundestag subset): [HuggingFace Dataset](https://huggingface.co/datasets/pdjohn/bundestag-causal-attribution)