The paper presents a method for relating brain structures to open-ended descriptions of cognition using text mining and statistical analysis. The goal is to establish a bidirectional mapping between brain activity and cognitive processes. The authors use the NeuroSynth database to extract cognitive terms and locations from neuroscience articles, enabling system-level studies of brain functional organization. They introduce a validation framework using information retrieval metrics to assess the accuracy of these mappings, focusing on relative frequencies of cognitive concepts. This approach allows for open-ended encoding and decoding of brain activity associated with cognitive terms. The experiments involve mapping activation coordinates from brain scans to text descriptions using the Cognitive Atlas ontology. The text is represented using Term Frequency - Inverse Document Frequency (TFIDF) features, and the brain activity is represented as a point in the vector space of brain voxels. Ridge Regression is used to predict one representation from the other, with encoding predicting brain activity from text and decoding predicting text from brain activity. The results show that both encoding and decoding accuracies are significantly above chance, demonstrating the effectiveness of the approach. The encoding results show that cognitive concepts, such as the language system, can be accurately captured from text descriptions. The decoding results show that known brain structures, such as the hippocampus for memory, can be reliably mapped from text descriptions. The authors also compare their results with those of the NeuroSynth model, finding that their approach performs significantly better in terms of decoding accuracy. The study highlights the potential of text mining and statistical analysis in understanding the relationship between brain activity and cognitive processes.The paper presents a method for relating brain structures to open-ended descriptions of cognition using text mining and statistical analysis. The goal is to establish a bidirectional mapping between brain activity and cognitive processes. The authors use the NeuroSynth database to extract cognitive terms and locations from neuroscience articles, enabling system-level studies of brain functional organization. They introduce a validation framework using information retrieval metrics to assess the accuracy of these mappings, focusing on relative frequencies of cognitive concepts. This approach allows for open-ended encoding and decoding of brain activity associated with cognitive terms. The experiments involve mapping activation coordinates from brain scans to text descriptions using the Cognitive Atlas ontology. The text is represented using Term Frequency - Inverse Document Frequency (TFIDF) features, and the brain activity is represented as a point in the vector space of brain voxels. Ridge Regression is used to predict one representation from the other, with encoding predicting brain activity from text and decoding predicting text from brain activity. The results show that both encoding and decoding accuracies are significantly above chance, demonstrating the effectiveness of the approach. The encoding results show that cognitive concepts, such as the language system, can be accurately captured from text descriptions. The decoding results show that known brain structures, such as the hippocampus for memory, can be reliably mapped from text descriptions. The authors also compare their results with those of the NeuroSynth model, finding that their approach performs significantly better in terms of decoding accuracy. The study highlights the potential of text mining and statistical analysis in understanding the relationship between brain activity and cognitive processes.