Pages Menu
Making Sense of Human - Human Conversation
Categories Menu

Data and SW



  • Syntactic parser (Macaon): This module provides sentence-level part-of-speech tagging and syntactic parses.
  • Semantic parser (Fastsem): This module performs sentence-level FrameNet semantic analysis of speech and social media text, given syntactic analyses provided by the syntactic parser.
  • Synopsis generator: This module generates synopses of speech conversations by recombining training data synopses and filling detected concept slot.
  • Sentiment analyser (SemEval2016): This module can predict the sentiment polarity of short text messages.
  • Abstractive cluster labeller: This rest service will label a textual piece with a label (abstractive) using DBpedia
  • Computation platform: This module helps with the integration of the various components as a general pipeline. It provides REST services and repository integration for other components.
  • Coreference resolver (BART): This module resolves coreferences in conversations.
  • Website parsers: These parsers can be used to crawl a range of websites for online conversations.
  • Discourse Parser (CoNLL2016): This system provides PDTB-style discourse parsing for written English text.
  • Agreement predictor (ADRian): This system provides agree/disagree classification in social media (used in social media prototype template-based summarisation, sense-eu).
  • Mood predictor (coMOOD): This system provides mood prediction (used in social media prototype template-based summarisation, sense-eu)
  • Repository: The conversation repository can store conversations and analyses generated by other modules. It is the central synchronization point of the pipeline.
  • Repository tools: Tools for easily accessing the repository from the command line
  • Comment clustering and summarization: This system can link comments in a conversation, cluster them and generate a summary from the clusters.
  • Event/sentiment detector (GATE): This system provides standard NLP, named-entity recognition, event detection, sentiment detection.
    • Models: English
    • License: LGPL (
      The tool includes two pipelines: one using the ARCOMEM sentiment detection software supplied complete, and one using the
      SentiStrength tool (better results) but it is missing the relevant jar which cannot be distributed by SENSEI.)
    • Open source: yes
    • Programming languages: java, groovy, jape
    • Repository integration: yes
    • Dependencies: GATE, ARCOMEM or SentiStrength
    • URL:
  • Social media eval prototype: This module contains the evaluation UI for the social media use case.
  • ACOF tool: This module provides an annotation interface for Agent Conversation Observation Form (ACOF) and support for the extrinsic evaluation.
Share this page
The research leading to these results has received funding from the European Union - Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 610916 – SENSEI.