Maki Miyake

Capturing the structures in association knowledge: application of network analyses to large-scale databases of Japanese word associations

Within the general enterprise of probing into the complexities of lexical knowledge, one particularly promising research focus is on word association knowledge. Given Deeses [1] and Cramers [2] convictions that word association closely mirror the structured patterns of relations that exist among concepts, as largely echoed Hirsts [3] more recent comments about the close relationships between lexicons and ontologies, as well as Firths [4] remarks about finding a words meaning in the company it keeps, efforts to capture and unravel the rich networks of associations that connect words together are likely to yield interesting insights into the nature of lexical knowledge. Adopting such an approach, this paper applies a range of network analysis techniques in order to investigate the characteristics of network representations of word association knowledge in Japanese. Specifically, two separate association networks are constructed from two different large-scale databases of Japanese word associations: the Associative Concept Dictionary (ACD) by Okamoto and Ishizaki [5] and the Japanese Word Association Database (JWAD) by Joyce [6] [7] [8]. Results of basic statistical analyses of the association networks indicate that both are scale-free with smallworld properties and that both exhibit hierarchical organization. As effective methods of discerning associative structures with networks, some graph clustering algorithms are also applied. In addition to the basic Markov Clustering algorithm proposed by van Dongen [9], the present study also employs a recently proposed combination of the enhanced Recurrent Markov Cluster algorithm (RMCL) [10] with an index of modularity [11]. Clustering results show that the RMCL and modularity combination provides effective control over cluster sizes. The results also demonstrate the effectiveness of graph clustering approaches to capturing the structures within large-scale association knowledge resources, such as the two constructed networks of Japanese word associations.

Using Graph Components Derived from an Associative Concept Dictionary to Predict fMRI Neural Activation Patterns that Represent the Meaning of Nouns.

In this study, we introduce an original distance definition for graphs, called the Markov-inverse-F measure (MiF). This measure enables the integration of classical graph theory indices with new knowledge pertaining to structural feature extraction from semantic networks. MiF improves the conventional Jaccard and/or Simpson indices, and reconciles both the geodesic information (random walk) and co-occurrence adjustment (degree balance and distribution). We measure the effectiveness of graph-based coefficients through the application of linguistic graph information for a neural activity recorded during conceptual processing in the human brain. Specifically, the MiF distance is computed between each of the nouns used in a previous neural experiment and each of the in-between words in a subgraph derived from the Edinburgh Word Association Thesaurus of English. From the MiF-based information matrix, a machine learning model can accurately obtain a scalar parameter that specifies the degree to which each voxel in (the MRI image of) the brain is activated by each word or each principal component of the intermediate semantic features. Furthermore, correlating the voxel information with the MiF-based principal components, a new computational neurolinguistics model with a network connectivity paradigm is created. This allows two dimensions of context space to be incorporated with both semantic and neural distributional representations.

Random Graph Model Simulations of Semantic Networks for Associative Concept Dictionaries

Word association data in dictionary form can be simulated through the combination of three components: a bipartite graph with an imbalance in set sizes; a scale-free graph of the Barabasi-Albert model; and a normal distribution connecting the two graphs. Such a model makes it possible to simulate the complex features in degree distributions and the interesting graph clustering results that are typically observed for real data.