The `pre-train, prompt, predict' paradigm of large language models (LLMs) has achieved remarkable success in open-domain question answering (OD-QA). However, few works explore this paradigm in multi-document question answering (MD-QA), a task demanding a thorough understanding of the logical associations among the contents and structures of documents. To fill this crucial gap, we propose a Knowledge Graph Prompting (KGP) method to formulate the right context in prompting LLMs for MD-QA, which consists of a graph construction module and a graph traversal module. For graph construction, we create a knowledge graph (KG) over multiple documents with nodes symbolizing passages or document structures (e.g., pages/tables), and edges denoting the semantic/lexical similarity between passages or document structural relations. For graph traversal, we design an LLM-based graph traversal agent that navigates across nodes and gathers supporting passages assisting LLMs in MD-QA. The constructed graph serves as the global ruler that regulates the transitional space among passages and reduces retrieval latency. Concurrently, the graph traversal agent acts as a local navigator that gathers pertinent context to progressively approach the question and guarantee retrieval quality. Extensive experiments underscore the efficacy of KGP for MD-QA, signifying the potential of leveraging graphs in enhancing the prompt design and retrieval augmented generation for LLMs. Our code: https://github.com/YuWVandy/KG-LLM-MDQA.

Given a large time-evolving graph, how can we model and characterize the temporal behaviors of individual nodes (and network states)? How can we model the behavioral transition patterns of nodes? We propose a temporal behavior model that captures the "roles" of nodes in the graph and how they evolve over time. The proposed dynamic behavioral mixed-membership model (DBMM) is scalable, fully automatic (no user-defined parameters), non-parametric/data-driven (no specific functional form or parameterization), interpretable (identifies explainable patterns), and flexible (applicable to dynamic and streaming networks). Moreover, the interpretable behavioral roles are generalizable and computationally efficient. We applied our model for (a) identifying patterns and trends of nodes and network states based on the temporal behavior, (b) predicting future structural changes, and (c) detecting unusual temporal behavior transitions. The experiments demonstrate the scalability, flexibility, and effectiveness of our model for identifying interesting patterns, detecting unusual structural transitions, and predicting the future structural changes of the network and individual nodes.

Viet Lai, Chien Nguyen, Nghia Ngo, Thuat Nguyen, Franck Dernoncourt, Ryan Rossi, Thien Nguyen. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2023.

This paper presents a general inductive graph representation learning framework called DeepGL for learning deep node and edge features that generalize across-networks. In particular, DeepGL begins by deriving a set of base features from the graph (e.g., graphlet features) and automatically learns a multi-layered hierarchical graph representation where each successive layer leverages the output from the previous layer to learn features of a higher-order. Contrary to previous work, DeepGL learns relational functions (each representing a feature) that naturally generalize across-networks and are therefore useful for graph-based transfer learning tasks. Moreover, DeepGL naturally supports attributed graphs, learns interpretable inductive graph representations, and is space-efficient (by learning sparse feature vectors). In addition, DeepGL is expressive, flexible with many interchangeable components, efficient with a time complexity of $O(|E|)$ , and scalable for large networks via an efficient parallel implementation. Compared with recent methods, DeepGL is (1) effective for across-network transfer learning tasks and large (attributed) graphs, (2) space-efficient requiring up to 6x less memory, (3) fast with up to 106x speedup in runtime performance, and (4) accurate with an average improvement in AUC of 20% or more on many learning tasks and across a wide variety of networks.

Random walks are at the heart of many existing node embedding and network representation learning methods. However, such methods have many limitations that arise from the use of traditional random walks, e.g., the embeddings resulting from these methods capture proximity (communities) among the vertices as opposed to structural similarity (roles). Furthermore, the embeddings are unable to transfer to new nodes and graphs as they are tied to node identity. To overcome these limitations, we introduce the Role2Vec framework based on the proposed notion of attributed random walks to learn structural role-based embeddings. Notably, the framework serves as a basis for generalizing any walk-based method. The Role2Vec framework enables these methods to be more widely applicable by learning inductive functions that capture the structural roles in the graph. Furthermore, the original methods are recovered as a special case of the framework when each vertex is mapped to its own function that uniquely identifies it. Finally, the Role2Vec framework is shown to be effective with an average AUC improvement of 17.8 percent for link prediction while requiring on average 853x less space than existing methods on a variety of graphs from different domains.

Jon Saad-Falcon, Joe Barrow, Alexa Siu, Ani Nenkova, Seunghyun Yoon, Ryan A. Rossi, Franck Dernoncourt. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track. 2024.