Learning Causality from Textual Data

Kira Radinsky, Sagie Davidovich and Shaul Markovitch. Learning Causality from Textual Data. In Proceedings of the IJCAI Workshop on Learning by Reading and its Applications in Intelligent Question-Answering, 363-367 Barcelona, Spain, 2011.

Abstract

We present a new methodology for modeling and predicting future events through machine learning and data mining techniques from textual data. Modeled events span across varied domains including politics, economy and society. The model employs human-style prediction techniques such as causality inference, generalization and projection based on past experience. For this purpose, we use news archives that date back 150 years as a vast source of text representing past-experiences and inference patterns. Empirical evaluation on real news articles shows that the ability of our algorithm to predict future events is similar to that of humans.

Keywords: Information Retrieval, Temporal Reasoning, Prediction

Secondary Keywords:

Online version:

Bibtex entry:

 @inproceedings{Radinsky:2011:LCT,
  Author = {Kira Radinsky and Sagie Davidovich and Shaul Markovitch},
  Title = {Learning Causality from Textual Data},
  Year = {2011},
  Booktitle = {Proceedings of the IJCAI Workshop on Learning by Reading and its Applications in Intelligent Question-Answering},
  Pages = {363--367},
  Address = {Barcelona, Spain},
  Url = {http://www.cs.technion.ac.il/~shaulm/papers/pdf/Radinsky-IJCAI2011ws.pdf},
  Keywords = {Information Retrieval, Temporal Reasoning, Prediction},
  Secondary-keywords = {Common-Sense Knowledge},
  Abstract = {
    We present a new methodology for modeling and predicting future
    events through machine learning and data mining techniques from
    textual data. Modeled events span across varied domains including
    politics, economy and society. The model employs human-style
    prediction techniques such as causality inference, generalization
    and projection based on past experience. For this purpose, we use
    news archives that date back 150 years as a vast source of text
    representing past-experiences and inference patterns. Empirical
    evaluation on real news articles shows that the ability of our
    algorithm to predict future events is similar to that of humans.
  }

  }