Introduction

KonText is an advanced corpus query interface for the Manatee-open corpus search engine. It builds on top of core server-side libraries from NoSketchEngine and both applications are data-compatible as well. The development is maintained by the Institute of the Czech National Corpus.

Features

new features

• fully editable query chain
  • any operation from a user defined sequence (e.g. query -> filter -> sample -> sorting) can be changed and the whole sequence is then re-executed.
• support for spoken corpora
  • defined concordance segments can be played back as audio
  • KWIC detail provides a custom rendering with easily distinguishable speeches
• support for user-defined line groups
  • user can define custom numeric tags attached to concordance lines, filter out other lines, review groups ratios
• improved subcorpus creation
  • user can easily examine corpus structure by selecting some text types and see how other text type attributes availability changed ("which publishers are there in case only fiction is selected?")
  • a custom text types ratio can be defined ("give me 20% fiction and 80% journalism")
  • a sub-corpus can be created by a custom CQL expression
  • subcorpora are backed up as CQL queries which makes further modification/restoring possible
• frequency distribution
  • 2-dimensional frequency distribution for both positional and structural attributes
  • result caching decreases time required to navigate between pages
  • on the multilevel frequency distribution page, starting word can be specified for multi-word KWICs
• persistent URLs for large queries - you can send a link to someone even if the query was in megabytes
• access to previous queries, named queries
• access to favorite corpora (subcorpora, aligned corpora)
• interactive PoS tag tool - in case of positional PoS tag formats an interactive tool can be used to write tag queries
• a concordance/frequency/collocation listing can be saved in Excel format (xlsx)
• a correct (i.e. the one calculating only with selected text types) i.p.m. can be calculated on-demand for ad-hoc subcorpora
• result shuffling can be pre-set
• less full page reloads

internal changes

• server-side rewritten as a WSGI application (Bonito-open is CGI-based)
• completely rewritten client-side code (React+Flux architecture, TypeScript + ES6, modularized)
• modular code design with dynamically loadable plug-ins providing custom functionality implementation (e.g. custom database adapters, authentication method, corpus listing widgets, HTTP session management)
• fully decoupled background concordance/frequency/collocation calculation based on the Celery task queue (alternatively, Python's multiprocessing package can be used)
• improved logging, error processing and debugging support
• improved code documentation

Requirements

• a WSGI-compatible server
  • recommended setup: Gunicorn + a reverse proxy (e.g. Nginx or Apache2)
  • supported setup: Apache2 with mod_wsgi
• Python 2.7 and:
  • Cheetah Template Engine
  • lxml library
  • werkzeug library (provides WSGI middleware)
  • PyICU library (optional but preferred)
  • markdown library (optional, for formatted corpora references)
  • openpyxl library (optional, for XLSX export)
• corpus search engine Manatee
  • versions from 2.83.3 to 2.150 are supported (the latest one is highly recommended); unless there is an incompatible change in Manatee, newer versions should work too
• a key-value storage
  • any custom implementation (Redis and SQLite backends are available by default)
• (optional) Celery task queue task queue for (asynchronous) background calculations and maintenance tasks

Build and installation

Please refer to the doc/INSTALL.md file for details.

Customization and contribution

Please refer to our Wiki.

Notable installations

• Institute of the Czech National Corpus
• LINDAT