DLM

Divergent Language Matrix (DLM) is designed to generate a lower-dimensional representation of complex, hierarchical text data, such as conversations. The algorithm preserves both semantic and structural relationships within the data, allowing for more efficient analysis and visualization. In the Divergent Language Matrix (DLM) framework, a conversation tree is formulated as a directed, acyclic graph where each node corresponds to a message in the conversation. Each message `t_i` is mathematically defined by a triplet `(d_i, s_i, c_i)`, such that: * x_coord (Depth): Represents the hierarchical level of a message. If a message is a direct reply to another message, it will be one level deeper (e.g., the original message is at depth 0, a reply to it is at depth 1, a reply to that reply is at depth 2, and so on). * y_coord (Order among siblings): Represents the order in which a message appears among its siblings. This is relevant when there are multiple replies (siblings) to a single message. It provides a sense of the sequence of the conversation. * z_coord (Homogeneity based on sibling count and similarity score): This is the most direct measure of homogeneity in the provided method.It serves as an essential indicator for both the structural and semantic relationships among messages at the same hierarchical level. The `z_coord` value is calculated differently depending on whether similarity scores are included.