Datasets:

havens2

/

naacl2022

like 0

[ "We", "choose", "the", "top", "nodes", "according", "to", "the", "value", "of", "LM-to-KG", "attention:" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "define", "a", "hyperparameter,", "the", "Retention", "ratio", "(0,", "1],", "which", "determines", "the", "number", "of", "nodes", "K", "to", "be", "retained." ]

[ 0, 0, 0, 0, 0, 3, 4, 5, 5, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0 ]

[ "Inspired", "by", "SAGPool", "(Lee", "et", "al.,", "2019),", "under", "the", "guidance", "of", "query,", "we", "retain", "relevant", "nodes", "and", "cut", "out", "irrelevant", "nodes", "according", "to", "the", "LM-to-KG", "attention." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "Equation", "10,", "the", "LM-to-KG", "attention", "value", "implies", "the", "importance", "of", "different", "nodes", "in", "the", "subgraph", "for", "question", "answering." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 12 ]

[ "ql", "i}", "{" ]

[ 0, 0, 0 ]

[ "M", "i=1", "will", "be", "input", "to", "the", "next", "l-th", "stacked", "JointLK", "layer", "of", "to", "continue", "participating", "in", "joint", "reasoning,", "and", "the", "V", "updated", "KG", "representation", "¯Xl", "=", "¯xl", "i=1", "will", "be", "|", "i}", "input", "to", "the", "next", "module", "of", "the", "current", "JointLK", "layer", "for", "pruning." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]

[ "The", "attended", "features", "are", "fused", "with", "the", "original", "features", "of", "the", "other", "modality", "by", "concatenation", "and", "then", "compressed", "to", "low-dimensional", "space", "by:" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "3.4", "e", "To", "reduce", "the", "gap", "of", "query", "and", "knowledge", "graph", "features,", "we", "fuse", "them", "in", "the", "joint", "reasoning", "module", "by", "the", "dense", "bidirectional", "attention", "mechanism", "that", "connects", "two", "encoding", "layers", "of", "query", "and" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "following", "joint", "reasoning", "module", "will", "further", "fuse", "i", "and", "ql", "xl", "to", "obtain", "their", "updated", "representations." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "ψ(eji,", "uj,", "ui)", "is", "the", "relation", "feature", "vector,", "where", "eji", "is", "a", "one-hot", "vector", "denoting", "the", "relation", "type", "of", "the", "edge", "(j,", "i)", "and", "uj,", "ui", "are", "one-hot", "vectors", "denoting", "the", "node", "types", "of", "j", "and", "i." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "D", "are", "where", "matrices", "Wq,", "Wk,", "Wv,", "Wo", "∈", "trainable", "parameters,", "Ni", "is", "the", "neighbor", "of", "node", "i.", "rji", "=" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "attended", "representations", "are", "computed", "as", "follows:" ]

[ 0, 0, 0, 0, 0, 0, 0 ]

[ "and", "also", "normalize", "Sl", "ij", "in", "column-wise", "to", "derive", "LM-to-KG", "attention", "maps", "on", "entities", "conditioned", "by", "each", "query", "token", "as" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "denote", "the", "inputs", "to", "the", "joint", "reasoning", "module", "in", "the", "l-st", "fusion", "layer", "by", "V", "1", "i=1." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "module", "takes", "the", "query", "and", "KG", "representations", "Q", "and", "X", "as", "inputs", "and", "then", "outputs", "their", "updated", "versions." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "knowledge", "graph", "and", "captures", "the", "fine-grained", "interplay", "between", "them." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "brevity,", "we", "formulate", "the", "entire", "computation", "in", "one", "layer", "as:" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Then,", "we", "apply", "GNN", "Layer", "to", "update", "node", "representation", "through", "iterative", "message", "passing", "between", "neighbors", "on", "the", "graph,", "while", "GNN", "is", "built", "on", "the", "RGAT", "(Wang", "et", "al.,", "2020a)", "and", "is", "a", "simplification", "of", "Yasunaga", "et", "al.", "(2021)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "First,", "We", "use", "the", "BERT", "model", "with", "average", "pooling", "to", "get", "the", "initial", "repRD." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "After", "obtaining", "token", "representations", "by", "the", "query", "encoder,", "we", "further", "model", "the", "subgraph", "to", "obtain", "entity", "representations." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "M", "i=1", "∈" ]

[ 0, 0, 0 ]

[ "The", "representations", "of", "tokens", "Q0", "=", "RD", "will", "be", "provided", "to", "the", "joint", "reasoning", "module", "for", "further", "interaction", "with", "the", "graph", "entities", "representations." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "RD", "is", "a", "linear", "transformation,", "where", "fs", ":", "RT", "and", "σ", "is", "the", "activation", "function." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Then", "we", "feed", "the", "representation", "of", "tokens", "into", "a", "non-linear", "layer", "so", "that", "the", "text", "representation", "space", "is", "aligned", "to", "the", "entity", "representation", "space:" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "follow", "baselines", "to", "use", "pre-trained", "language", "M", "wi}", "i=1", "(question", "and", "models", "to", "encode", "the", "query", "{", "M", "i=1:", "choice)", "into", "a", "sequence", "of", "vectors" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "describe", "the", "detailed", "extraction", "process", "in", "Appendix", "A." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "E", "V", "is", "the", "set", "of", "edges", "that", "connect", "nodes", "in", "V", ",", "where", "R", "is", "a", "set", "of", "relations", "types." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Here", "V", "is", "a", "subset", "of", "entity", "nodes", "retrieved", "from", "the", "external", "KG." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "(V,", "R)", "with", "the", "guidance", "of", "question", "and", "choice." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "extract", "from", "the", "external", "KG", "a", "subgraph", "g", "=" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "general,", "questions", "do", "not", "contain", "any", "reference", "to", "answer", "choices,", "so", "the", "external", "knowledge", "graph", "provides", "the", "necessary", "background", "knowledge." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Given", "a", "commonsense", "question", "q", "and", "a", "set", "of", "answer", "choices", "a1,", "a2,", "...,", "an}", ",", "our", "task", "is", "to", "measure", "the", "plausi{", "bility", "score", "between", "q", "and", "each", "answer", "choice", "a", "then", "select", "the", "answer", "with", "the", "highest", "plausibility", "score." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "CSQA", "task", "in", "this", "paper", "is", "a", "multiple-choice", "problem", "with", "some", "answer", "choices." ]

[ 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "3.1", "Task", "Definition" ]

[ 0, 0, 0 ]

[ "After", "N", "layers", "of", "iteration,", "the", "query", "representation", "and", "the", "trimmed", "graph", "representation", "are", "used", "to", "predict", "the", "answer", "(§", "3.6)." ]

[ 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "weights", "and", "finally", "retains", "the", "most", "relevant", "nodes", "(§", "3.5)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "LM-to-KG", "attention", "weights", "generated", "in", "reasoning", "represents", "the", "global", "importance", "of", "each", "node", "in", "the", "graph,", "so", "the", "dynamic", "pruning", "module", "prunes", "the", "graph", "layer", "by", "layer", "according", "to", "this" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "Joint", "Reasoning", "Module", "receives", "these", "two", "modalities’", "representations", "and", "then", "apply", "dense", "bidirectional", "attention", "to", "make", "information", "fusion", "and", "representation", "update", "for", "each", "token", "and", "node", "(§", "3.4)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "use", "a", "pre-trained", "language", "model", "to", "learn", "the", "query", "representation", "(§", "3.2),", "and", "use", "the", "GNN", "layer", "to", "learn", "the", "graph", "representation", "(§", "3.3)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "model", "is", "mainly", "composed", "of", "four", "parts:", "query", "encoder,", "GNN", "layer,", "joint", "reasoning", "module", "and", "dynamic", "pruning", "module,", "of", "which", "the", "latter", "three", "form", "a", "stack", "of", "N", "identical", "layers." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0 ]

[ "JointLK", "takes", "the", "query", "and", "the", "retrieved", "knowledge", "subgraph", "as", "input,", "and", "outputs", "a", "real", "value", "as", "the", "correctness", "score", "of", "the", "answer." ]

[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "model", "framework", "is", "shown", "in", "Figure", "2." ]

[ 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "this", "section,", "we", "introduce", "the", "task", "definition", "(§", "3.1)", "and", "our", "JointLK", "model." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]

[ "3", "Methodology" ]

[ 0, 0 ]

[ "Furthermore,", "we", "gradually", "prune", "the", "KG", "size", "in", "each", "stacked", "model", "layer", "under", "the", "guidance", "of", "attention", "weights", "generated", "in", "the", "interactions,", "making", "the", "reasoning", "path", "transparent", "and", "interpretable." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "grained", "interaction", "between", "any", "token", "in", "question", "and", "any", "entity", "in", "KG", "through", "dense", "bidirectional", "attention,", "and", "perform", "multi-step", "joint", "reasoning", "by", "stacking", "several", "interaction", "layers." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "JointLK", "mainly", "consists", "of", "four", "modules", "the", "Query", "Encoder,", "the", "Graph", "Layer,", "the", "Joint", "Reasoning", "Module", "and", "the", "Dynamic", "Pruning", "Module,", "of", "which", "the", "latter", "three", "form", "a", "stack", "of", "N", "identical", "layers." ]

[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0 ]

[ "Compared", "with", "prior", "works,", "we", "retain", "the", "individual", "structure", "of", "both", "modalities,", "consider", "fine" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "it", "pools", "the", "representation", "of", "the", "question", "context", "into", "a", "single", "node,", "which", "limits", "the", "updating", "of", "the", "text", "representation", "and", "fine-grained", "interaction", "between", "LM", "and", "GNN." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Recently,", "QA-GNN", "(Yasunaga", "et", "al.,", "2021)", "explicitly", "views", "the", "QA", "context", "as", "an", "additional", "node,", "connects", "it", "and", "KG", "to", "form", "a", "joint", "graph,", "and", "mutually", "updates", "their", "representations", "through", "graphbased", "message", "passing." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Although", "this", "method", "can", "retain", "the", "original", "information", "of", "question", "context", "and", "KGs,", "the", "limited", "interaction", "will", "affect", "the", "flow", "of", "information", "between", "the", "two", "modalities,", "so", "we", "mainly", "improve", "on", "this", "point." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Other", "works", "(Lin", "et", "al.,", "2019;", "Feng", "et", "al.,", "2020;", "Yan", "et", "al.,", "2021)", "use", "LM", "and", "GNN", "to", "model", "the", "two", "modalities", "separately,", "and", "perform", "shallow", "interactions", "in", "the", "latter", "model", "stage,", "such", "as", "attentive", "pooling", "or", "simple", "concatenation", "of", "the", "two", "modal", "representations." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "the", "original", "structural/textual", "information", "will", "inevitably", "be", "lost", "during", "the", "conversion", "process." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Some", "works", "(Lv", "et", "al.,", "2020;", "Bian", "et", "al.,", "2021;", "Xu", "et", "al.,", "2021)", "unify", "the", "two", "modalities", "during", "model", "input,", "such", "as", "transforming", "structured", "knowledge", "into", "plain", "text", "through", "templates", "or", "transforming", "question", "context", "into", "structured", "graphs." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Due", "to", "the", "heterogeneity", "between", "structured", "knowledge", "and", "unstructured", "text", "questions,", "there", "are", "currently", "two", "main", "research", "methods." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Therefore,", "many", "works", "obtain", "the", "required", "knowledge", "from", "external", "sources", "(e.g.,", "KGs,", "corpus)", "to", "augment", "CSQA", "models." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0 ]

[ "Commonsense", "question", "answering", "is", "challenging", "because", "the", "required", "commonsense", "knowledge", "is", "rarely", "given", "in", "the", "context", "of", "questions", "and", "answer", "choices", "or", "encoded", "in", "the", "parameters", "of", "pre-trained", "LMs." ]

[ 11, 12, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "2", "Related", "Work" ]

[ 0, 0, 0 ]

[ "Furthermore,", "through", "the", "multi-layer", "fusion", "of", "these", "two", "modalities,", "JointLK", "exhibits", "strong", "performance", "over", "previous", "state-of-the-art", "LM+KG", "methods", "in", "performing", "complex", "reasoning,", "such", "as", "solving", "questions", "with", "negation", "and", "complex", "questions", "with", "more", "entities." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "•", "Experimental", "results", "show", "that", "JointLK", "is", "superior", "to", "current", "LM+KG", "methods,", "and", "the", "refined", "evidence", "is", "interpretable." ]

[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "nodes", "at", "each", "JointLK", "layer", "to", "ensure", "that", "the", "model", "reasons", "correctly", "with", "complete", "and", "appropriate", "evidence." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "It", "uses", "dense", "bidirectional", "attention", "to", "simultaneously", "update", "query-aware", "knowledge", "graph", "representation", "and", "knowledgeaware", "query", "representation,", "bridging", "the", "gap", "between", "the", "two", "information", "modalities." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "propose", "JointLK,", "a", "novel", "model", "that", "supports", "multi-step", "joint", "reasoning", "between", "LM", "and", "KG." ]

[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "•" ]

[ 0 ]

[ "In", "summary,", "our", "contributions", "are", "three-fold:" ]

[ 0, 0, 0, 0, 0, 0 ]

[ "Multiple", "JointLK", "layers", "are", "stacked", "to", "form", "a", "hierarchy", "that", "supports", "multi-step", "interactions", "and", "recursive", "pruning." ]

[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Guided", "by", "the", "attention", "generated", "in", "the", "interaction", "process,", "the", "dynamic", "pruning", "module", "deletes", "irrelevant", "nodes", "to", "make", "the", "model", "reason", "along", "the", "correct", "knowledge", "path." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Then", "we", "design", "a", "joint", "reasoning", "module", "to", "generate", "fine-grained", "bidirectional", "attention", "maps", "between", "each", "question", "token", "and", "each", "KG", "node", "to", "fuse", "the", "information", "from", "each", "modality", "to", "the", "other." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Specifically,", "given", "a", "question", "and", "retrieved", "subgraphs,", "JointLK", "first", "obtain", "the", "representations", "of", "the", "two", "modalities", "by", "using", "an", "LM", "encoder", "and", "a", "GNN", "encoder", "respectively." ]

[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Based", "on", "the", "above", "consideration,", "we", "propose", "JointLK,", "a", "model", "that", "performs", "the", "fine-grained", "modal", "fusion", "and", "multi-layer", "joint", "reasoning", "between", "the", "language", "model", "and", "the", "knowledge", "graph", "(see", "Figure", "2)." ]

[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "argue", "that", "the", "limited", "interaction", "between", "the", "two", "modalities", "is", "the", "main", "bottleneck", "that", "may", "prevent", "the", "model", "from", "understanding", "the", "complex", "question-knowledge", "relations", "necessary", "to", "answer", "the", "question", "correctly." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Specifically,", "existing", "LM+KG", "methods", "(Lin", "et", "al.,", "2019;", "Feng", "et", "al.,", "2020)", "model", "question", "context", "and", "knowledge", "subgraphs", "in", "isolation", "by", "LMs", "and", "GNNs,", "and", "perform", "only", "one", "interaction", "in", "a", "shallow", "manner", "to", "fuse", "their", "representations", "at", "the", "output", "for", "prediction." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Second,", "there", "are", "limited", "interactions", "between", "language", "representation", "and", "knowledge", "graph", "representation." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "As", "the", "example", "in", "Figure", "1,", "some", "graph", "nodes", "such", "as", "“wood”,", "“burn”,", "and", "“gas”,", "although", "related", "to", "some", "entities", "in", "the", "questions", "and", "choice,", "can", "mislead", "the", "global", "understanding", "of", "the", "question." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "simple", "string", "matching", "or", "semantic", "matching,", "in", "order", "to", "retrieve", "sufficient", "relevant", "knowledge,", "noise", "knowledge", "graph", "nodes", "will", "inevitably", "be", "included", "(Lin", "et", "al.,", "2019;", "Yasunaga", "et", "al.,", "2021)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Whether", "through" ]

[ 0, 0 ]

[ "First,", "the", "retrieved", "knowledge", "subgraph", "contains", "many", "noisy", "nodes." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "these", "approaches", "have", "two", "main", "issues." ]

[ 0, 0, 0, 0, 0, 0, 0 ]

[ "Second,", "the", "retrieved", "subgraphs", "are", "modeled", "by", "a", "well-designed", "graph", "neural", "network", "module", "(Lin", "et", "al.,", "2019;", "Feng", "et", "al.,", "2020;", "Yasunaga", "et", "al.,", "2021)", "to", "perform", "reasoning", "over", "knowledge", "graphs." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "First,", "the", "knowledge", "subgraphs", "or", "paths", "related", "to", "a", "given", "question", "are", "retrieved", "by", "string", "matching", "or", "semantic", "similarity;", "such", "retrieved", "structured", "information", "indicates", "the", "relation", "between", "concepts", "or", "implies", "the", "process", "of", "multi-hop", "reasoning." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Related", "methods", "usually", "follow", "a", "retrieval-and-modeling", "paradigm." ]

[ 0, 0, 0, 0, 0, 0, 0 ]

[ "An", "extensive", "research", "path", "is", "to", "elaborately", "design", "graph", "neural", "networks", "(GNNs)", "(Scarselli", "et", "al.,", "2008)", "to", "perform", "reasoning", "over", "explicit", "structural", "common", "sense", "knowledge", "from", "external", "knowledge", "bases", "(Vrandeˇci´c", "and", "Krötzsch,", "2014;", "Speer", "et", "al.,", "2017)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "knowledge", "is", "self-evident", "to", "humans", "and", "is", "rarely", "expressed", "clearly", "in", "natural", "language", "(Gunning,", "2018),", "which", "makes", "it", "difficult", "for", "LMs", "to", "learn", "commonsense", "knowledge", "from", "the", "pre-training", "text", "corpus", "alone." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Nevertheless,", "commonsense" ]

[ 0, 0 ]

[ "Recently,", "large", "pre-trained", "language", "models", "(LMs)", "have", "achieved", "remarkable", "success", "in", "many", "QA", "tasks", "and", "appear", "to", "use", "implicit", "(factual)", "knowledge", "encoded", "in", "their", "model", "parameters", "during", "fine-tuning", "(Liu", "et", "al.,", "2019;", "Raffel", "et", "al.,", "2020)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Commonsense", "question", "answering", "(CSQA)", "requires", "systems", "to", "acquire", "different", "types", "of", "commonsense", "knowledge", "and", "reasoning", "skills,", "which", "is", "normal", "for", "humans,", "but", "challenging", "for", "machines", "(Talmor", "et", "al.,", "2019)." ]

[ 11, 12, 12, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Introduction" ]

[ 0 ]

[ "We", "evaluate", "JointLK", "on", "the", "CommonsenseQA", "and", "OpenBookQA", "datasets,", "and", "demonstrate", "its", "improvements", "to", "the", "existing", "LM", "and", "LM+KG", "models,", "as", "well", "as", "its", "capability", "to", "perform", "interpretable", "reasoning1." ]

[ 0, 0, 1, 0, 0, 13, 0, 13, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 12 ]

[ "Then,", "the", "dynamic", "pruning", "module", "uses", "the", "attention", "weights", "generated", "by", "joint", "reasoning", "to", "prune", "irrelevant", "KG", "nodes", "recursively." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Specifically,", "JointLK", "performs", "joint", "reasoning", "between", "LM", "and", "GNN", "through", "a", "novel", "dense", "bidirectional", "attention", "module,", "in", "which", "each", "question", "token", "attends", "on", "KG", "nodes", "and", "each", "KG", "node", "attends", "on", "question", "tokens,", "and", "the", "two", "modal", "representations", "fuse", "and", "update", "mutually", "by", "multi-step", "interactions." ]

[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "this", "paper,", "we", "propose", "a", "novel", "model,", "JointLK,", "which", "solves", "the", "above", "limitations", "through", "the", "joint", "reasoning", "of", "LM", "and", "GNN", "and", "the", "dynamic", "KGs", "pruning", "mechanism." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "they", "ignore", "(i)", "the", "effectively", "fusing", "and", "reasoning", "over", "question", "context", "representations", "and", "the", "KG", "representations,", "and", "(ii)", "automatically", "selecting", "relevant", "nodes", "from", "the", "noisy", "KGs", "during", "reasoning." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Existing", "KG-augmented", "models", "for", "commonsense", "question", "answering", "primarily", "focus", "on", "designing", "elaborate", "Graph", "Neural", "Networks", "(GNNs)", "to", "model", "knowledge", "graphs", "(KGs)." ]

[ 0, 0, 0, 0, 11, 12, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Abstract" ]

[ 0 ]

[ "JointLK:", "Joint", "Reasoning", "with", "Language", "Models", "and", "Knowledge", "Graphs", "for", "Commonsense", "Question", "Answering" ]

[ 1, 1, 2, 2, 2, 2, 2, 2, 2, 0, 11, 12, 12 ]

[ "We", "find", "the", "contrastive", "loss", "particularly", "helpful", "when", "the", "amount", "of", "speech", "data", "is", "extremely", "small,", "like", "only" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "a", "fair", "comparison,", "we", "use", "the", "same", "MT", "pre-trained", "Transformer", "module", "as", "in", "the", "main", "paper." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "the", "experiment,", "we", "reduce", "the", "labeled", "ST", "data", "to", "1,", "10,", "and", "100", "hours,", "corresponding", "to", "sentence", "counts", "of", "about", "500,", "5k,", "and", "50k", "sentences." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Here,", "we", "simulate", "the", "scenario", "with", "plenty", "of", "MT", "and", "speech", "data", "and", "limited", "ST", "triple-labeled", "data,", "and", "does", "ConST", "have", "the", "ability", "of", "low-resource", "learning?" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]

[ "The", "experiments", "in", "the", "main", "paper", "show", "that", "our", "model", "can", "perform", "well", "by", "introducing", "external", "MT", "data", "pre-training." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]