This repository contains the script that allows for the automatic morphosyntactic annotation of Russian sentences containing a verb of propositional attitude with a subordinate clause.

This script uses spacy library for morphosyntactic parsing together with pymorphy2 library for morpphological parsing. pymorphy often show better performance on moorphology than spacy does. In the script, they are often used to complement each other.

sample_input.xlsx is a sample file for the script to annotate. It contains the following columns (bold are given).

• Source: database from which the context was extracted
• Verb: target verb
• Embedding: presence of negation
• PreContext: two sentences before the target
• Target: target sentence
• PostContext: two sentences after the target
• MatTense: tense of the main predicate
• MatSubjPers: person value of the subject of the main predicate
• MatSubjNum: number values of the subject of the main predicate
• MatAspect: aspect of the main predicate
• SubTense: tense of the subordinate predicate
• SubSubjPers: person value of the subject of the subordinate predicate
• SubSubjNum: number values of the subject of the subordinate predicate
• SubAspect: aspect of the subordinate predicate
• Conjunction: conjunction connecting main and subordinate clauses

The final dataframe contains the same columns, but all filled.

The parsing is rule-based, however due to incorrect syntactic annotation, the final table if full of mistakes. It is possible to replace spacy model with any other model and get better or worse results.