---
title: Multi-perspective Knowledge Extraction
url: https://www.emergentmind.com/topics/multi-perspective-knowledge-extraction-mke
type: topic
---

# Multi-perspective Knowledge Extraction

Multi-perspective Knowledge Extraction (MKE) denotes a class of methods that extract, generate, retrieve, or verify knowledge from multiple predefined perspectives and then reintegrate those perspective-specific signals into a downstream model. In recent arXiv work, the term appears most explicitly in description-based knowledge graph completion, where the task is to produce richer textual information from three perspectives—reasoning, explanation, and summarization—by querying a large language model and incorporating the outputs back into a knowledge graph; closely related formulations use multi-view retrieval in retrieval-augmented generation and multi-axial knowledge with evidence verification in ICD coding for Chinese electronic medical records [2403.01972] [2404.12879] [2502.14916].

## 1. Conceptual scope and recurring design pattern

Across these formulations, the central assumption is that a single representation is insufficient in knowledge-dense settings. In "Multi-perspective Improvement of Knowledge Graph Completion with Large Language Models" [2403.01972], the three perspectives are reasoning-based entity expansion, explanation-based relation understanding, and summarization-based structure extraction. In "Unlocking Multi-View Insights in Knowledge-Dense Retrieval-Augmented Generation" [2404.12879], the relevant construct is a set of predefined domain viewpoints or “professional perspectives,” such as “Medical History,” “Symptoms,” “Application of Law,” and “Focus of Dispute.” In "MKE-Coder: Multi-Axial Knowledge with Evidence Verification in ICD Coding for Chinese EMRs" [2502.14916], ICD codes are decomposed into four “axes”: Etiology axis, Anatomical site axis, Pathology axis, and Clinical manifestation axis.

These systems differ in target task, but they share a common workflow. First, a task-specific object is decomposed into multiple views, perspectives, or axes. Second, each perspective is used to produce an intermediate representation: rewritten queries in MVRAG, augmented entity or relation text and SameAs edges in MPIKGC, and axis-specific keyword sets plus evidence sentences in MKE-Coder. Third, the perspective-specific outputs are fused back into a conventional learner: a reader LLM in RAG, a description-based KGC scorer, or a prompt-tuned T5 verifier. This suggests that MKE is best understood as a family of augmentation-and-fusion strategies rather than a single canonical architecture.

| Formulation | Perspective set | Reintegration target |
|---|---|---|
| MPIKGC | reasoning, explanation, summarization | description-based KGC model |
| MVRAG | predefined domain viewpoints | retrieval and final generation |
| MKE-Coder | four coding axes | evidence-supported ICD code inference |

## 2. Formalization in description-based knowledge graph completion

In the knowledge-graph setting, the formal object is
$$
G = \{ (h,r,t,d_h,d_r,d_t) \mid h,t\in\mathcal{E},\ r\in\mathcal{R},\ d_h,d_r,d_t\in\mathcal{D} \}.
$$
A description-based KGC model learns an encoder $\Phi$ that maps textual inputs to embeddings,
$$
e_h = \Phi(d_h \parallel name(h)), \quad v_r = \Phi(d_r \parallel name(r)), \quad e_t = \Phi(d_t \parallel name(t)),
$$
together with a scoring function $f : \mathbb{R}^d \times \mathbb{R}^d \times \mathbb{R}^d \to \mathbb{R}$, and a typical margin-ranking loss
$$
L = \sum_{(h,r,t)\in\mathcal{T}^+} \sum_{(h',r,t')\in\mathcal{T}^-} \max(0, \gamma + f(e_{h'},v_r,e_{t'}) - f(e_h,v_r,e_t)).
$$
The MKE task is defined as: “Given a (possibly sparse) $G$ with minimal or missing $d_h, d_r, d_t$, produce richer textual information from three perspectives—reasoning (entity expansion), explanation (relation understanding), summarization (structure extraction)—by querying a large language model (LLM). Incorporate the LLM outputs back into $G$ so as to minimize $L$ and improve link-prediction and triplet-classification accuracy.” [2403.01972]

The reasoning module, MPIKGC-E, expands sparse entity descriptions through a Chain-of-Thought-style prompt: “Please provide all information about {Entity Name}. Give the rationale before answering:”. The generated paragraph $d_h^{LLM}$ is concatenated to the original $d_h$ or the entity name alone before encoding via $\Phi$. The explanation module, MPIKGC-R, augments relation names with three one-sentence explanations: a global prompt, a local triplet rephrasing prompt, and a reverse-direction prompt. The integration step is
$$
d_r \leftarrow name(r) \parallel expl\_global(r) \parallel expl\_local(r) \parallel expl\_reverse(r).
$$
The summarization module, MPIKGC-S, extracts five representative keywords from an entity description, computes
$$
s(h,t) = \frac{|k_h \cap k_t|}{\min(|k_h|, |k_t|)},
$$
selects the top-$k$ highest-scoring pairs, and adds triples $(h,SameAs,t)$ together with self-loops $(e,SameAs,e)$ to $\mathcal{T}^+$. After augmentation, the loss remains a margin-ranking or contrastive loss over $\mathcal{T}^+_{aug}$.

This formulation is notable for being model-agnostic at the KGC layer. The backbones evaluated include KG-BERT, SimKGC, LMKE, and CSProm-KG, while the LLMs used for generation include LLaMA-2-7B-chat, ChatGLM2-6B, ChatGPT (gpt-3.5-turbo), and GPT-4. The perspective modules therefore function as textual and structural augmentation components rather than replacements for the scoring model.

## 3. Multi-view retrieval as perspective-conditioned knowledge extraction

In retrieval-augmented generation, the multi-perspective extraction problem is framed as intention-aware query rewriting and weighted retrieval. MVRAG extends a standard RAG pipeline by inserting two additional “multi-view” stages—Intention Recognition and Query Rewriting—before retrieval, and a weighted re-ranking step after retrieval. The pipeline is: input query $q$ to Intention Recognition to produce a Perspective Vector $V_q$; for each non-zero entry in $V_q$, produce a subquery $q_i^{(v)}$; retrieve a document set $R_i$ per subquery; re-rank all retrieved documents using both similarity and perspective weights; and concatenate the top documents with the original query into a prompt $\mathcal{P}$ for the reader LLM [2404.12879].

The perspective vector is
$$
V_q = [\,w_1,\;w_2,\;\dots,\;w_n\,],
$$
with
$$
w_i =
\begin{cases}
LLM_{\mathrm{score}(q,p_i)}, & \text{if } LLM_{\mathrm{score}(q,p_i)}>\theta,\\
0, & \text{otherwise.}
\end{cases}
$$
Here $P=\{p_1,\dots,p_n\}$ is the set of predefined domain viewpoints, $LLM_{\mathrm{score}(q,p_i)}$ is a relevance score, and $\theta$ is a small threshold such as $0.1$. For each $i$ with $w_i>0$, the viewpoint-aware subquery is
$$
q_i^{(v)} = C_i = \mathrm{Rewriter}(q,p_i).
$$
For each non-zero perspective weight $w_i$, the retriever returns
$$
R_i = \bigl\{d_{ij}\mid d_{ij}\text{ ranks among top }k_i\text{ by }\cossim(d_{ij},C_i)\bigr\},
$$
with $k_i \propto w_i$, for example $k_i=\lceil k_{\max}\cdot w_i\rceil$. Each document is then re-scored as
$$
\mathrm{Rel}(d_{ij}) = \cossim(d_{ij},C_i)\times w_i,
$$
and the top $M$ pooled documents are concatenated to form
$$
\mathcal{P} = q \;\Vert\; d'_1 \;\Vert\; d'_2 \;\Vert\,\dots\Vert\; d'_M.
$$

A distinctive feature of this formulation is that no additional learned embeddings are required: the viewpoints $p_i$ enter as natural-language descriptors, and MVRAG uses two LLM calls per perspective, one for scoring and one for rewriting. The paper notes that one could optionally embed $p_i$ and learn a small adapter, but the framework treats LLMs as black boxes. In knowledge-dense domains such as law and medicine, this design makes the perspective labels directly inspectable and links retrieved documents to explicit professional viewpoints.

## 4. Multi-axial extraction and evidence verification for ICD coding

MKE-Coder applies the MKE idea to automatic ICD coding for Chinese EMRs by decomposing each candidate code $c$ into four mappings: Etiology axis $(A_1)$, Anatomical site axis $(A_2)$, Pathology axis $(A_3)$, and Clinical manifestation axis $(A_4)$. The multi-axis knowledge is represented as
$$
K(c) = \{ K_1(c), K_2(c), K_3(c), K_4(c) \},
$$
where each $K_j(c)$ is the small vocabulary of keywords or phrases defining axis $A_j$. In practice, these are obtained by splitting the code string according to the official ICD indexing rules, looking up each segment in an ICD-code→description dictionary, extracting noun phrases or headwords, and expanding each keyword set with clinical synonyms via UMLS or manual synonym lists. For example, code H31.403 is decomposed into Etiology $\{$“inflammatory,” “degenerative”$\}$, Anatomical site $\{$“retina,” “choroid”$\}$, Pathology $\{$“morphological change,” “atrophy”$\}$, and Clinical manifestation $\{$“vision loss,” “floaters”$\}$ [2502.14916].

The end-to-end system begins with preprocessing: Chinese word segmentation with jieba, removal of rare symbols, normalization of digits and units, sentence splitting on punctuation marks, and section tagging by source field. Candidate code pre-selection uses
$$
C_d = Top_n\ Sort_{(descending)} \{ Sim(d,c) : c \in C \},
$$
with
$$
Sim(d,c) = w_1\cdot ED(d,c) + w_2\cdot TF(d,c) + w_3\cdot FEA(d,c),
$$
where $w_1=0.35$, $w_2=0.35$, and $w_3=0.30$. For evidence retrieval, the overall axis keyword set is
$$
K_{all}(c) = \bigcup_{j=1}^{4} [ K_j(c) \cup Syn(K_j(c)) ].
$$
Within each likely section $\ell \in L(c)$, sentences are scored by
$$
g(\ell,s;c) = | s \cap K_{all}(c) |,
$$
and the best sentence is
$$
s_{\ell}^* = \arg\max_{s \in S(\ell)} g(\ell,s;c).
$$
Sentences are greedily accumulated into $S_{temp}(c)$ until every keyword in $K_{all}(c)$ has appeared or the section list is exhausted.

Evidence filtering uses a Chinese SimBERT model fine-tuned in two stages. At inference time,
$$
E_s = SimBERT(s), \qquad E_k = SimBERT(k),
$$
and
$$
Score(s,k) = cos(E_s,E_k)=\frac{E_s\cdot E_k}{\|E_s\|\cdot\|E_k\|}.
$$
The final evidence set is
$$
E(c) = \{ s \in S_{temp}(c) : Score(s,k_s) \ge T \},
$$
with threshold $T=0.65$. Code verification is then converted into a fill-in-the-blank problem using a T5 encoder and soft prompt tokens:
$$
T(c) = [CLS]\ s^1 \dots s^Q\ [soft_1]\dots[soft_m]\ w_1\dots w_P\ [soft_1]\dots[soft_m]\ [MASK].
$$
The verbalizer is $y\in\{0,1\}\mapsto\{$“no”, “yes”$\}$, the prediction is
$$
p(y|T(c)) = p([MASK] = Verbalizer(y)\mid T(c)),
$$
and the objective is
$$
\mathcal{L} = - \sum_{(c,y)} \log p(y|T(c)).
$$
Only the soft tokens are learned; all T5 weights remain frozen.

## 5. Empirical findings and ablation evidence

In multi-view RAG, the reported gains are large on both legal and medical case retrieval. On LeCaRDv2, bge-m3 rises from 3.13% Recall@100, 0.94% Precision@100, and 1.44% F1 to 16.53%, 4.96%, and 7.63% with MVRAG. On PMC-Patients, gte-large rises from 1.60% Recall@100, 0.11% Precision@100, and 0.20% F1 to 13.05%, 2.13%, and 3.67%. The paper states that all models saw consistent, large relative gains, often 4–10× improvements in Recall@100. An ablation that removes one perspective at a time shows that omitting “Fundamental Facts” in the legal task drops Recall@100 from 16.53% to 12.11% [2404.12879].

In description-based KGC, the gains are smaller in absolute magnitude but consistent across backbones and tasks. On FB15k-237 with CSProm-KG, the original model reports MR 188, MRR 35.23, H@1 26.05, H@3 38.72, and H@10 53.57; +MPIKGC-E gives MR 195, MRR 35.51, H@1 26.38, H@3 38.96, and H@10 53.74; +MPIKGC-R gives MR 192, MRR 35.38, H@1 26.29, H@3 38.83, and H@10 53.50; +MPIKGC-S gives MR 179, MRR 35.95, H@1 26.71, H@3 39.52, and H@10 54.30. On triplet classification with KG-BERT on FB13, the original accuracy is 84.74%, +MPIKGC-E is 86.29%, +MPIKGC-R is 84.51%, and +MPIKGC-S is 85.35%. The LMKE ablation on FB15k-237 shows that +S gives the largest single-perspective net gain on Hits@10, while the pairwise combination R+S achieves the best MRR and H@1 [2403.01972].

In ICD coding, MKE-Coder is evaluated on a 70 K-record Chinese EMR set with Macro Precision, Recall, F1, Accuracy, and Precision@5. The reported state-of-the-art results are 72.61% macro-F1, 74.32% accuracy, and 92.73% P@5. Interpretability is quantified by expert review, which reports 50% fully supported and 46% partially supported evidence. The practical evaluation within simulated real coding scenarios states that the method significantly aids coders in enhancing both their coding accuracy and speed [2502.14916].

## 6. Interpretability, limitations, and points of clarification

A central motivation for MKE is interpretability. In MVRAG, multi-view retrieval “surfaces” which domain perspectives contributed to each retrieved document via the weight $w_i$, making it clear, for instance, that a case was retrieved chiefly because it matched “Symptoms” rather than “Laboratory Data.” The Huntington’s disease case study illustrates the point: vanilla RAG retrieved Vitamin B12–deficiency articles because of superficial symptom match, whereas MVRAG’s symptoms and history perspectives directed retrieval toward neurodegeneration and genetic-testing literature, avoiding misdiagnosis [2404.12879].

In MPIKGC, interpretability is distributed across prompts and augmentation channels rather than a single explanation module. Entity expansion yields explicit fact-rich paragraphs; relation understanding provides global, local, and reverse-direction semantics; and structure extraction adds SameAs edges through a transparent keyword-overlap rule. A possible source of confusion is the status of those SameAs edges. Within MPIKGC-S, they are generated by selecting top-$k$ entity pairs according to keyword-set similarity and then treating the resulting triples as additional positive facts appended to the training set. The example pairing of “Michael Bay” and “Transformers: Dark of the Moon” follows this scoring procedure, not a separate entity-resolution algorithm [2403.01972].

In MKE-Coder, interpretability is tied to evidence verification rather than only to candidate ranking. Each supported code is returned with its evidence set $E(c)$, where evidence is first retrieved by axis-keyword coverage and then filtered by Clinical-SimBERT. The final T5 module verifies whether all the axis knowledge associated with the candidate code is supported by evidence and provides recommendations accordingly. This suggests a stricter notion of support than retrieval-only systems, because the framework explicitly asks whether a candidate code is supported by evidence rather than merely similar to the diagnosis text [2502.14916].

Taken together, these works indicate that multi-perspective extraction is most useful when sparse inputs, ambiguous labels, or heterogeneous domain evidence make one-view modeling unreliable. They also indicate that perspective choice is task-dependent: professional viewpoints for RAG, reasoning/explanation/summarization for KGC, and four coding axes for ICD inference. The current literature therefore treats “perspective” less as a fixed ontology than as a structured decomposition chosen to expose otherwise missing or weakly represented knowledge.

Source: https://www.emergentmind.com/topics/multi-perspective-knowledge-extraction-mke