While at the morphemic level, every word of the sentence is assigned a lemma and a tag and at the analytical level, every word corresponds to a node in the analytical tree, with the appropriate analytical function, the tectogrammatical level has no such close relation to the surface form of a sentence.
The relation between the nodes of the tectogrammatical and analytical level (which is generally of the type M:N, the options 0:N and M:0 included) is captured by the atree.rf attribute of the technical root of the tectogrammatical tree and by the a attribute (the value of which is a structure of the lex.rf and a/aux.rf attributes) with other nodes.
The atree.rf attribute. The atree.rf attribute refers trivially to the technical root of the analytical tree corresponding to the given tectogrammatical tree. It contains the identifier of the technical root node of the corresponding analytical tree (see Table 2.1, "Values of the atree.rf attribute").
Table 2.1. Values of the atree.rf attribute
| PML reference |
the identifier of the technical root node of the corresponding analytical tree |
The a/lex.rf and a/aux.rf attributes. The a/lex.rf attribute contains the identifier of the node at the analytical level from which the tectogrammatical node got its lexical meaning (or its biggest part). The a/aux.rf attribute contains the list of the identifiers of all analytical nodes that influence in some way or other the value of the functor, subfunctor or grammatemes of the tectogrammatical node (these are mostly analytical nodes representing so called function words like prepositions, conjunctions, auxiliaries and anaphoric words). See Table 2.2, "Values of the a/lex.rf attribute" and Table 2.3, "Values of the a/aux.rf attribute".
Table 2.2. Values of the a/lex.rf attribute
| PML reference |
the identifier of the node at the analytical level from which the tectogrammatical node got its lexical meaning (or its biggest part) |
Table 2.3. Values of the a/aux.rf attribute
| a list every element of which is a PML reference |
identifiers of the analytical nodes that influence in some way or other the value of the functor, subfunctor or grammatemes of the tectogrammatical node |
The values of these attributes conform to the following rules:
If the tectogrammatical node has no analytical counterpart, both attributes are empty.
Example:
Dovolil mu odejít. {#Cor.ACT} (=He let him go)
A new node is added to the tectogrammatical tree, a node with the #Cor t-lemma and the ACT functor, which has no counterpart at the tectogrammatical structure. Both the a/lex.rf and a/aux.rf attributes are empty.
If a tectogrammatical node (other than a newly established one with one of the t-lemmas #Forn, #Idph, #EmpVerb or #EmpNoun) corresponds to exactly one analytical node, than the a/lex.rf attribute contains the reference to the (analytical) node and the a/aux.rf is empty.
Example:
Premiér zahájil schůzi. (=The Prime Minister opened the meeting)
Each of these words is represented by a single node in both the analytical and tectogrammatical trees. The a/lex.rf attribute of each of the tectogrammatical nodes contains the identifier of the corresponding analytical node. The a/aux.rf attribute is empty with all of the nodes.
If a tectogrammatical node (other than a newly established one with one of the t-lemmas: #Forn, #Idph, #EmpVerb or #EmpNoun) corresponds to more analytical nodes, then the a/lex.rf attribute contains the reference to the node from which the tectogrammatical node got its lexical meaning (or its bigger part) and the a/aux.rf attribute contains the list of references to the other analytical nodes, which mostly represent function words (prepositions, subordinating conjunctions, auxiliaries etc.).
Example:
Odešli s tím, že už by nemuseli nikdy přijít. (=lit. (They) left with that that (they) already would not_have_to never come)
In the tectogrammatical tree representing this sentence, the (single) node with the t-lemma přijít corresponds to six analytical nodes, namely: s, tím, že, by, nemuseli, přijít. The a/lex.rf attribute of the tectogrammatical node contains the identifier of the corresponding analytical node for přijít. The a/aux.rf attribute contains the list of the identifiers of the remaining analytical nodes: the supporting expression (the prepositional phrase) and the subordinating conjunction have influence on the functor of the tectogrammatical node, the auxiliary (the conditional form of the verb být) and the modal verb have influence on the values of the grammatemes.
With newly established nodes with the t-lemma #EmpVerb or #EmpNoun, the a/lex.rf attribute is always empty since the full verb they represent is not expressed at the surface level. If no function word is expressed at the surface level either, the a/aux.rf attribute is also empty; otherwise it contains the list of references to the relevant function words (e.g. auxiliaries that are part of a complex verb form where the full verb is not expressed).
Example:
{#EmpVerb.COND} Pokud ano, dejte nám vědět. (=If yes, let us know)
A new node is added to the tectogrammatical tree, in the position of (the predicate of) the dependent clause, a node with the #EmpVerb t-lemma and the COND functor, which corresponds to no node (no full verb) in the analytical tree. The functor value is influenced by the subordinating conjunction pokud. The a/lex.rf attribute is therefore empty, the a/aux.rf contains the identifier of the analytical node representing pokud.
With newly established nodes with the t-lemmas #Idph and #Forn, which serve for putting parts of identifying and foreign-language expressions together, into a single list (nodetype=list), the a/lex.rf attribute is always empty. If a foreign-language or identifying expression is syntactically combined with one or more function words at the analytical level, then the references to these function words are contained in the a/aux.rf attribute.
Example:
{#Forn.DIR3} Pojede na Pikes Peak. (=He is going to Pikes Peak)
A new node is added to the tectogrammatical tree, in order to represent the foreign-language expression (as such), a node with the #Forn t-lemma and the DIR3 functor, which corresponds to no node in the analytical tree. The functor value is influenced by the preposition na. The a/lex.rf attribute is therefore empty, the a/aux.rf contains the identifier of the analytical node representing na.
NB! Units of the lower levels do not have to have counterparts at the tectogrammatical level. Thus, various graphic symbols are ignored as well as the reflexive se when part of a reflexive passive and constructions with dispositional modality.
Copied nodes. b) and c) also concern copied nodes. These are nodes representing lexical items present at the tectogrammatical level several times, which are however expressed only once at the surface level because their other occurrences were elided. The a/lex.rf and a/aux.rf attributes of the copied nodes contain the identifiers of the analytical nodes for the words present at the surface level that are relevant for the copied node, i.e. that influence its t-lemma, functor and other attributes. Cf.:
Řešení je rozděleno na dvě části. První {část.PAT} se bude realizovat na našem pracovišti. (=The solution is divided into two parts. The first one will be carried out in our department.)
The prepositional phrase na části (in the first sentence) is represented by a single node in the tectogrammatical tree: the a/lex.rf attribute contains the identifier of the analytical node for the noun část, the a/aux.rf attribute contains the identifier of the analytical node for the preposition na. The tectogrammatical tree for the second sentence contains a node copied from the first sentence (the one representing the prepositional phrase na části), in the position of the elided Patient. The a/lex.rf attribute of this copied node will again contain the identifier of the analytical node for the noun část; the a/aux.rf attribute will be empty, though; the preposition na has no influence on the functor of the copied node.
For more on ellipsis, see Section 12, "Ellipsis".
Node with the #PersPron t-lemma. a), b) and c) also concern nodes with the #PersPron t-lemma. Nodes with the #PersPron t-lemma do not have to correspond to any analytical nodes; then the rules in a) are followed. Cf.:
{#PersPron.ACT} Přišel. (=He came)
The node with the #PersPron t-lemma corresponds to no analytical node. Both the a/lex.rf and a/aux.rf attributes are empty.
When a node with the #PersPron t-lemma represents a personal or reflexive pronoun present at the surface level, the rules in b) or c) are followed. The a/lex.rf attribute contains the identifier of the analytical node representing the pronoun. The a/aux.rf attribute contains the identifiers of (possible) function words. Cf.:
On přišel. (=He came)
The personal pronoun on, even when expressed, is represented by a node with the #PersPron t-lemma. The a/lex.rf attribute of the this node contains the identifier of the analytical node representing the pronoun on. The a/aux.rf attribute is empty.
Přišel pro něj. (=He came for him)
The prepositional phrase pro něj is represented by a node with the #PersPron t-lemma. The a/lex.rf attribute of the this node contains the identifier of the analytical node representing the pronoun něj. The a/aux.rf attribute contains the identifier of the analytical node for the preposition pro.
Node with the #Rcp t-lemma. a), b) and c) also concern nodes with the #Rcp t-lemma. Nodes with the #Rcp t-lemma do not have to correspond to any analytical nodes; then, the rules in a) are followed. Cf.:
{#Rcp.ADDR} Státy spolu obchodují. (=The states trade with each other)
The reciprocal relation in this sentence is captured by adding a new node into the structure, one with the #Rcp t-lemma. A node with the #Rcp t-lemma corresponds to no analytical node. Both the a/lex.rf and a/aux.rf attributes are empty.
The rules in b) or c) are applied in those cases when the surface form contains some formal indication of reciprocity. The a/lex.rf attribute contains the identifier of the analytical node representing the reciprocal pronoun se. The a/aux.rf attribute contains the identifier of the analytical node representing a preposition - if present. Cf.:
Pavel a Martin se potkali. {#Rcp.PAT} (=lit. Pavel and Martin REFL met)
The reciprocal relation in this sentence is captured by adding a new node to the structure, one with the #Rcp t-lemma. The a/lex.rf attribute of the this node contains the identifier of the analytical node representing se. The a/aux.rf attribute is empty.
Martin a Radek se mezi sebou domluvili. {#Rcp.ADDR} (=Martin and Radek made an arrangement; lit. between themselves)
The reciprocal relation in this sentence is captured by adding a new node into the structure, one with the #Rcp t-lemma. The a/lex.rf attribute of the this node contains the identifier of the analytical node representing se. The a/aux.rf attribute contains the identifier of the analytical node for the preposition mezi.
For more on reciprocity, see Section 2.4.2, "Reciprocity").
!!! References to prepositions, subordinating conjunctions and supporting expressions were checked after the annotation was finished, so they can be considered reliable. The situation regarding references to auxiliary and modal verbs is different: they were not checked. Also the a/aux.rf attribute values with copied nodes were checked only partially. All attribute values remained without any change (which we suppose is correct) with those copied nodes the governing node of which has the same t-lemma as the copied node. It was also tested whether the copied nodes contain references to only one conjunction or preposition in the a/aux.rf attribute.