Skip to content

Treesearch

High-performance dependency treebank querying for corpus linguistics.

Quick Start

pip install treesearch

import treesearch as ts

# Find all passive constructions
query = """
MATCH {
    V [upos="VERB"];
    V -[aux:pass]-> _;
    V -[nsubj:pass]-> Subj;
}
"""

for tree, match in ts.search("corpus/*.conllu", query):
    verb = tree.word(match["V"])
    subj = tree.word(match["Subj"])
    print(f"{subj.form} was {verb.form}: {tree.sentence_text}")

Features

  • Query language for structural patterns (nodes, edges, precedence, negation)
  • Regular expressions for flexible pattern matching with automatic anchoring
  • Exhaustive search finds all matches using CSP solving
  • Automatic parallelism for multi-file processing
  • Memory efficient streaming with string interning
  • Transparent gzip support

Quick Examples

# Exact match
ts.search("*.conllu", 'MATCH { V [lemma="run"]; }')

# Regex: words ending in -ing
ts.search("*.conllu", 'MATCH { V [form=/.*ing/]; }')

# Regex: VERB or AUX
ts.search("*.conllu", 'MATCH { V [upos=/VERB|AUX/]; }')

# Complex: progressive construction
query = """
MATCH {
    Aux [lemma=/be.*/];      # be, is, was, etc.
    V [form=/.*ing/];        # -ing form
    Aux -[aux]-> V;
}
"""

Documentation

License

MIT