Cognitive Mapping

The data consists of text and relations, which have three parts (Concept1, Explanation and Concept2). Each of the three parts correspond to a multi-word phrase in a text.

All relations and text can be easily read into a Jupyter notebook.

The goal is to identify the three parts of the relation in a text automatically.

Example:

'3-2: [concept Giving to the ECB the ultimate responsibility for supervision of banks in the euro area concept] [explanation will decisively contribute to increase explanation] [concept confidence between the banks concept] [explanation and in this way increase explanation] [concept the financial stability in the euro area concept]. The euro area governments and the European institutions, including naturally the European Commission and the ECB, will do whatever is necessary to secure the financial stability of the euro area.\n'

Plan

1. machine learning of paragraphs: do they contain a causal relation or not
2. find phrases of relations in text: either concepts, explanations or not present in a relation
3. identify relations based on recognized concept phrases and explanation phrases

We need a tagger or a entity recognition program, for example transformers, huggingface/bert: https://github.com/huggingface/transformers , or Spacy

Contributing

Contributons are welcome, see CONTRIBUTING.

Citation

If you want to cite this software, please use the metdata in CITATION.cff.

References

Hosseini, M.J., Chambers, N., Reddy, S., Holt, X R., Cohen, S B., Johnson, M., & Steedman, M. (2018). Learning Typed Entailment Graphs with Global Soft Constraints, Transactions of the Association for Computational Linguistics. Sizov, G. & Ozturk, P. (2013).

Zornitsa Kozareva, Irina Matveeva, Gabor Melli, Vivi Nastase Zornitsa Kozareva, Irina Matveeva, Gabor Melli, Vivi Nastase (2013). Automatic Extraction of Reasoning Chains from Textual Reports. Proceedings of TextGraphs-8 Workshop “Graph-based Methods for Natural Language Processing”, Empirical Methods in Natural Language Processing.

Noah Jadallah (2021). Cause-Effect Detection for Software Requirements Based on Token Classification with BERTCause-Effect Detection for Software Requirements Based on Token Classification with BERT. Seminar Natural Language Processing for Software Engineering, Winter-term 2020/2021, Technical University Munich.

Erik Tjong Kim Sang and Katja Hofmann (2009). Lexical Patterns or Dependency Patterns: Which Is Better for Hypernym Extraction? In: Proceedings of CoNLL-2009, Boulder, CO, USA, 2009, pages 174-182.