Neurolinguistic approach to natural language processing with applications to medical text analysis

Abstract

Understanding written or spoken language presumably involves spreading neural activation in the brain. This process may be approximated by spreading activation in semantic networks, providing enhanced representations that involve concepts not found directly in the text. The approximation of this process is of great practical and theoretical interest. Although activations of neural circuits involved in representation of words rapidly change in time snapshots of these activations spreading through associative networks may be captured in a vector model. Concepts of similar type activate larger clusters of neurons, priming areas in the left and right hemisphere. Analysis of recent brain imaging experiments shows the importance of the right hemisphere non-verbal clusterization. Medical ontologies enable development of a large-scale practical algorithm to re-create pathways of spreading neural activations. First concepts of specific semantic type are identified in the text, and then all related concepts of the same type are added to the text, providing expanded representations. To avoid rapid growth of the extended feature space after each step only the most useful features that increase document clusterization are retained. Short hospital discharge summaries are used to illustrate how this process works on a real, very noisy data. Expanded texts show significantly improved clustering and may be classified with much higher accuracy. Although better approximations to the spreading of neural activations may be devised a practical approach presented in this paper helps to discover pathways used by the brain to process specific concepts, and may be used in large-scale applications.

Neurocognitive Approach to Clustering of PubMed Query Results

“…A combination of all four indexes can be normalized between 0 and 4 so that 4 means perfect agreement between all the measures as for the number of clusters. Prior work that used automated priming approach showed usefulness of combined indices in discovery of interesting clusters in patient discharge summaries [1,13,14].…”

“…Multiple evidence for priming effects comes from psychology and neuroimaging. For that reason presented here technique for information retrieval is neurocognitively inspired [1,13,14].…”

“…Instead of boosting productivity increasingly large proportion of time is spent on searching and evaluating results. An ideal search and presentation system should be matched to the neurocognitive mechanisms responsible for understanding information [1]. Visualization of clusters of documents that are semantically related should reflect relations between configurations of neural activations in the brain of an expert.…”

“…An interesting research direction is based on knowledge-based clustering techniques [9]. In this paper neurocognitive inspirations [1] are used to create semantic graphs that represent PubMed search results. In the next section the outline of this approach is presented, followed by the description of methods and experiments on a restricted PubMed domain.…”

Optimal Support Features for Meta-Learning

“…The raw features given in the dataset description are used to create a large set of enhanced or hidden features. The topic of feature generation has received recently more attention in analysis of sequences and images, where graphical models known as Conditional Random Fields became popular [23], generating for natural text analysis sometimes millions of low-level features [24]. Attempts at meta-learning on the ensemble level lead to very rough granularity of the existing models and knowledge [25], thus exploring only a small subspace of all possible models, as it is done in the multistrategy learning [26].…”