In today’s information society, mechanisms for automatic natural language processing are widely integrated, evident in activities such as internet text searches, automatic translation, user-friendly interface development, and more. Much of this relates to computational linguistics, whose theories and methods have gradually become essential not only for academic research but also for everyday technology. Consequently, the influence of computational linguistics on models of communication and information sharing among people is increasing significantly. Computational linguistics is used to analyze phenomena at various levels, including phonetics and phonology, morphology, syntax, semantics, and pragmatics. However, it is most heavily applied during syntactic analysis, which involves the structural analysis of sentences. Similarly, corpus linguistics, as a method for studying large, computationally accessible corpora, often uses treebanks to examine syntactic phenomena. The use of treebanks has already gained a significant role in foreign language learning and translation education, with one example being the development of practical manuals, which are frequently based specifically on findings from such research. Our goal is to demonstrate how representing syntactic structures and processing them with computer programs is achievable. According to linguistic tradition, there are two developmental trends related to syntactic structures (Carstensen et al., 2010, p. 233). One views syntactic structures as connections between words; this involves dependency and determination syntax (Dependenz- und Determinationsyntax). L. Tesnière is a key figure in dependency theory. During dependency analysis, the relationships between individual words in a sentence are illustrated. The second approach involves the syntax of constituent structure (phrase structure syntax), which, in addition to words, also includes complex units called constituents or phrases, and the links between them. It is believed that sentences consist of hierarchically organized sub-units—sequences of interconnected words, known as constituents. Notably, the research of Leonard Bloomfield (1933) and American structuralists is essential in this area. Within the scope of the research, it became possible to annotate several German and Georgian sentences to identify the features of how verbal phrases function in both languages. However, in the next phase, the annotation of additional German and Georgian sentences, as well as their structural alignment, is also planned. The goal of this research was to prepare the ground for creating Georgian-German parallel treebanks. A parallel treebank is a parallel corpus of texts where sentences are annotated at various levels and typically presented as tree structures. A parallel corpus includes texts and their translations. Due to the lack of sufficient texts of this kind for some language pairs, similar text types in the respective languages are also considered parallel texts. They are used in areas such as machine translation, lexicography, and linguistic research aimed at understanding the relationships between structurally different languages. Syntactic Annotation/Software Tool "Synpathy" This work analyzes verbal phrases based on syntactically annotated German and Georgian sentences from a specific text type. Syntactic annotation includes parts of speech, morphological features, and information about syntactic dependencies. It is assumed that each sentence has a head, which functions as a node in the tree structure and is positioned above the entire sequence of words. Words and phrases that are semantically linked into a single constituent should be subordinate to one node. The tree branches in such a way that the lines describing the structure of the tree, known as branches or “twigs”, do not intersect each other. The symbols representing word form categories are called nodes. Tree leaves are terminal nodes that correspond to only one word. All other types of nodes except terminal ones, that is, final nodes that are not subject to further constituent decomposition, are called non-terminal. All nodes taken together that stand directly above words are referred to as pre-terminal nodes (Kapanadze & Steiner, 2012). For the syntactic annotation of a monolingual corpus, we utilized the software package Synpathy, developed by the Max Planck Institute for Psycholinguistics in the Netherlands. It is based on the SyntaxViewer software package created for the TIGER research project. Monolingual text is provided to Synpathy as files in TIGER-XML format. Monolingual treebanks converted into TIGER-XML format create powerful, database-like graph structures. In the TIGER-XML tree graph, each "leaf" (token) and each node (linguistic constituent) has a unique identifier. For each sentence, the corresponding software file in TIGER-XML format includes a "header" and a "body." The "header" contains meta-information, such as the corpus name, date, author, and explanations for the tags used, among other details. The “body” part includes: (1) terminal units (tree “leaves”) represented by individual words, part-of-speech tags, and associated features as attributes; (2) non-terminal units that describe corresponding nodes and the connecting lines between nodes. The annotation process involves two steps: grammatical information is assigned to each word, including its part-of-speech tag from the relevant tagset, and the syntactic structure is encoded as a tree-like graph with nodes and edges. Words in sentences are tagged using POS. Tree nodes are marked based on syntactic categories of constituents. Line marking is done through daughter functions in relation to corresponding parent nodes. The result of syntactic annotation is provided in TIGER XML format, which has two parts. The header section describes the corpus and includes some meta-information. The document body contains the definitions of corpus graphs and their annotations. The corpus ID (identification indicator) must be unique across the entire indexed corpus, meaning it should not be duplicated. Here, at the lowest level nodes, we encounter symbols like N (noun) and V (verb), while at the next, more dominant level nodes, we find more complex symbols. These are symbols not for individual words but for groups of words, such as: DP - for determiner phrase, NP - for nominal phrase, VP - for verbal phrase. Within the scope of the mentioned research, in addition to theoretical foundations, we examined the construction of sentence derivational trees based on X' rules. Specifically, we describe a particular sentence using program code (script) in the Notepad++ tool. In the first stage, we identify the words and punctuation marks of this sentence using terminal symbols. After finishing the program code, we save the document and verify it with the software tool “Synpathy.” If the code is correct, an X’ derivational tree will automatically open. Phrase Structure In the early stages of developing the theory of generative syntax, scholars aimed to establish all syntactic components of grammar based on phrase structures. The fundamental parts of syntax are represented by words that are linked through specific features of lexical categories. A sentence is not made up directly of individual words but of groups of words that are closely connected and together form phrases. Phrase categories are named according to the word that functions as the central element in the phrase, that is, its root or head (Schiller et al., 1999). For example: • Nominal phrases (NP): The head of this phrase is usually a word that inflects, typically a noun or pronoun. For example, NP - der Effektivzinssatz, Verzugszinsen. • Prepositional phrases (PP): The head of this phrase is a preposition. PP - for example, Einwilligung zur Übermittlung von Daten an die SCHUFA. • Verbal phrases (VP): The head of this phrase is represented by either a full verb or copula. VP - for example, Einwilligung widerrufen, Zinsen berechnen. • Adjectival phrases (AdjP): The head of this phrase is represented by an adjective. AdjP - for example, laufendes Konto, festes Sollzins. • Adverbial phrases (AdvP): The head of this phrase is represented by an adverb. AdvP - for example, unter Berücksichtigung aller Umstände des Einzelfalls, unter Abwägung der beiderseitigen Interessen, auf Verlangen des Kreditnehmers. According to I. According to Pafel (2011), we can also consider sentences as groups of words, that is, as phrases. However, in the literature, phrases are used in a narrow sense, where a sentence is not considered a phrase (Pafel, 2011, p. 24). According to L. Lemnitzer (2010), endocentric and exocentric phrases are distinguished. In an endocentric phrase, there is a phrase head that determines the category of the phrase, such as a noun in a nominal phrase. In contrast, an exocentric phrase functions as a single unit where multiple phrases are combined, like a sentence that encompasses nominal and verbal phrases. In this case, the so-called parent node is of a different type than its daughter nodes. Verbal Phrase Structure Phrases establish syntactic categories at both the word and sentence levels. They represent immediate constituents. For example, a verbal phrase results from expanding a lexical category verb that has a head. Verbal phrases contain a single finite verb or copula as the main element. Finite verbs are essential in phrases, especially when they require complements to fill specific roles. These complements must have a certain form, which is why they are called dative, genitive, accusative, and prepositional complements or objects. The nominative object is traditionally seen as the subject, playing a special role in various grammatical models. It is considered relatively independent from the verb and is often placed outside the verbal phrase. There are many reasons for this. The nominative or subject case is viewed as an ungoverned case by the verb, while the verb governs accusative, dative, and genitive cases. Even in ancient grammars, the nominative is referred to as casus rectus (correct, direct case), unlike casus obliqui (incorrect, oblique case). The strong connection between the subject and the verb is also evidenced by the fact that they cannot be combined with non-finite verbs. Additionally, the subject influences finite verbs, which must agree with it in person and number. The existence of verbal phrases in German is highly debated, especially regarding what should be included in a verbal phrase. Some believe the subject does not belong to the verbal phrase; others think it does, in which case the verbal phrase essentially encompasses the entire sentence, except for initial conjunctions and certain adverbs (Pittner & Berman, 2013). Every verb forms its own independent phrase. Phrases containing finite verbs are labeled as S (sentence), while non-finite phrases are labeled as VP (verb phrase). Verb phrases and clauses that serve as complements of nominal phrases are marked as OC (clausal object). It is important to note that if, as a result of grammaticalization, a specific connection between a noun and a verb becomes clear (e.g., in Verbindung setzen, in Anspruch nehmen, zur Verfügung stellen, in Verzug sein), then the analysis of noun arguments with verbal and clausal roles as OC (clausal object) is also possible. Verb phrases and clauses that are generally labeled as MO (Modifier/Modifikator) (e.g., clauses with um-zu) are labeled in nominal phrases as MNR (Modifier of NP to the Right/postnominal modifier). It should be noted that the subject is marked as a dependent of the finite verb. An entire clause can also be marked as OC. Annotation of Infinitive Constructions with the zu Particle The infinitive and the 'zu' particle are initially linked by a VZ phrase (PM is a morphological particle), and this phrase serves as the head of the subsequent main verb phrase. It should be noted that all components (syntactic complements/additions) and adjuncts of the verb (in generative literature, attributes are called adjuncts, i.e., so-called modifiers) attach to the VP node and not to VZ. The syntactic tree adequately reflects the functor-argument dependency relationship between verbs. The annotation of infinitive constructions with the zu particle can be represented as follows: Figure I. Tree Structure of German Sentence The same sentence tree structure is shown in the following schema in Georgian, which is especially important for this particular type of text, specifically for banking contracts. Figure II. Tree Structure of Georgian Sentence Collocational Verb Constructions (CVC) Collocational verb constructions (funktionales Verbgefüge/collocational verb construction) refer to combinations of the so-called semantically bleached full verbs and prepositional phrases. The semantic meaning in this case is carried by the prepositional phrase rather than the verb itself - as in [in Verbindung] cvc setzen könnte. Collocational verb constructions are characterized by the possibility of replacing them with a single verb (substitution). For example: in Verbindung setzen – verbinden. The main prepositions are zu or in. Replacing the noun involved in the construction is generally not possible without changing the meaning. Here, we deal with a small number of verbs that have weak meanings, mainly directional or locative (stellen, setzen, bringen, geraten, kommen, stehen, ...). In collocational verb constructions, the main group is labeled as CVC, although they can also be represented in a complex structural form: [[mit der Bank] op in Verbindung] cvc setzen could be. In collocational verb constructions, noun arguments with verbal and clausal values are marked as OC (clausal object): Figure III. Tree Structure of German Sentence Passive Voice Constructions In passive voice constructions, it is important to note that, in exceptional cases, the annotated structure does not align with the semantic argumentative structure. The subject, which semantically belongs to the verb's arguments (geführt), like all subjects, attaches to the finite auxiliary verb of the passive (wird). When annotating the passivized logical subject, we use the marker SBP (passivized subject). It is also possible for the clause argument (OCs) to be passivized; in this case, they can be annotated as subjects. Figure IV. Tree Structure of German Sentence Figure V. Tree Structure of Georgian Sentence The annotation of the stative passive (der Zustandspassiv) is done in the same way as the processual passive (der Vorgangspassiv), with the only difference being that the participle, or the verb phrase that dominates the participle, is annotated not as OC but as PD. In the stative passive, the participle is tagged as VVPP. The participle is mainly tagged as VVPP, except for lexicalized participles that are used as adjectives. For example: bekannt, begabt. The meaning of these adjectives has nothing in common with the original form of the verb, so they are assigned the PoS-Tag ADJD. The stative passive and copular sentences (Kopulasatz) have much in common, and their distinction is not so simple (e.g., Die Tür ist geschlossen). Participles in the stative passive have, on one hand, adjectival properties and can undergo lexicalization as adjectives (but not idiomatization). In other cases, there is a productive connection with the verbal paradigm, which becomes apparent through the werden-passive (e.g., Die Tür wird geschlossen), which results from the connection of the edge marker PD with the PoS-Label VVPP. Within the scope of the present article, we will not dwell on those phrases, as they are also considered verbal participles and are understood as part of the perfect with the verb sein, i.e., where one verb carries meaning. Conclusion For today’s world, particularly for European countries, work on monolingual and parallel treebanks is highly relevant. Corpus linguists strive to include as many languages as possible in the database and, accordingly, obtain more parallel corpora. In conditions of growing globalization, overcoming language barriers occupies a central place. At the same time, the use of computer is increasing in society. There is ever-growing demand for translation and computer services, which creates an ideal environment for the use of natural language machine translation systems. When performing translation, knowledge of content-related and cultural relationships connected to both linguistic spaces is extremely important, which in the case of literary texts is complemented by the diversity of individual creative thinking, so that adequate transfer of text characteristics to the target language is possible. From the day of the computer's invention, attempts have been made to integrate translation tasks into its other functions. Thus emerged the interdisciplinary field of machine translation, which combines the influence of linguistics, mathematics, informatics, and translation studies. Its goal is to approach the quality of translation performed by humans. Recently, instead of professional translators, the so-called CAT (computer-aided translation) technologies are used, with the help of which different types of multilingual corpora are created (terminology databases, aligned texts) and which can be used for subsequent translations. For example, in compiling technical equipment manuals, controlled language is partially incorporated so that final automatic processing and refinement of created documents – as well as machine translation – can be achieved. Against the background of ever-increasing technologization, the demand for machine translation use is extremely high. Today, compared to manual translation, the efficiency of the machine translation process has increased by 50%. However, keeping pace with these trends is possible through quite difficult and labor-intensive, step-by-step work. The present article shows exactly one such attempt.