Document (#44173)

Author

Luo, L.
Ju, J.
Li, Y.-F.
Haffari, G.
Xiong, B.
Pan, S.

Title

ChatRule: mining logical rules with large language models for knowledge graph reasoning

Source

https://www.researchgate.net/publication/373685576_ChatRule_Mining_Logical_Rules_with_Large_Language_Models_for_Knowledge_Graph_Reasoning

Year

2023

Abstract

Logical rules are essential for uncovering the logical connections between relations, which could improve the reasoning performance and provide interpretable results on knowledge graphs (KGs). Although there have been many efforts to mine meaningful logical rules over KGs, existing methods suffer from the computationally intensive searches over the rule space and a lack of scalability for large-scale KGs. Besides, they often ignore the semantics of relations which is crucial for uncovering logical connections. Recently, large language models (LLMs) have shown impressive performance in the field of natural language processing and various applications, owing to their emergent ability and generalizability. In this paper, we propose a novel framework, ChatRule, unleashing the power of large language models for mining logical rules over knowledge graphs. Specifically, the framework is initiated with an LLM-based rule generator, leveraging both the semantic and structural information of KGs to prompt LLMs to generate logical rules. To refine the generated rules, a rule ranking module estimates the rule quality by incorporating facts from existing KGs. Last, a rule validator harnesses the reasoning ability of LLMs to validate the logical correctness of ranked rules through chain-of-thought reasoning. ChatRule is evaluated on four large-scale KGs, w.r.t. different rule quality metrics and downstream tasks, showing the effectiveness and scalability of our method.

Theme

Computerlinguistik

Object

ChatRule

Similar documents (author)

1. Xiong, L.J.: On the compiling of cataloguing rules for Chinese document(s) (1997) 6.19

   6.190705 = sum of:
     6.190705 = weight(author_txt:xiong in 3199) [ClassicSimilarity], result of:
       6.190705 = fieldWeight in 3199, product of:
         1.0 = tf(freq=1.0), with freq of:
           1.0 = termFreq=1.0
         9.905128 = idf(docFreq=5, maxDocs=44218)
         0.625 = fieldNorm(doc=3199)
   

2. Xiong, C.: Knowledge based text representations for information retrieval (2016) 6.19

   6.190705 = sum of:
     6.190705 = weight(author_txt:xiong in 5820) [ClassicSimilarity], result of:
       6.190705 = fieldWeight in 5820, product of:
         1.0 = tf(freq=1.0), with freq of:
           1.0 = termFreq=1.0
         9.905128 = idf(docFreq=5, maxDocs=44218)
         0.625 = fieldNorm(doc=5820)
   

3. Xiong, S.; Ji, D.: Query-focused multi-document summarization using hypergraph-based ranking (2016) 4.95

   4.952564 = sum of:
     4.952564 = weight(author_txt:xiong in 2972) [ClassicSimilarity], result of:
       4.952564 = fieldWeight in 2972, product of:
         1.0 = tf(freq=1.0), with freq of:
           1.0 = termFreq=1.0
         9.905128 = idf(docFreq=5, maxDocs=44218)
         0.5 = fieldNorm(doc=2972)
   

4. Zeng, Q.; Yu, M.; Yu, W.; Xiong, J.; Shi, Y.; Jiang, M.: Faceted hierarchy : a new graph type to organize scientific concepts and a construction method (2019) 2.48

   2.476282 = sum of:
     2.476282 = weight(author_txt:xiong in 400) [ClassicSimilarity], result of:
       2.476282 = fieldWeight in 400, product of:
         1.0 = tf(freq=1.0), with freq of:
           1.0 = termFreq=1.0
         9.905128 = idf(docFreq=5, maxDocs=44218)
         0.25 = fieldNorm(doc=400)
   

Similar documents (content)

1. Panyr, J.: Information retrieval techniques in rule-based expert systems (1991) 0.21

   0.20763563 = sum of:
     0.20763563 = product of:
       0.8651485 = sum of:
         0.03453557 = weight(abstract_txt:knowledge in 3036) [ClassicSimilarity], result of:
           0.03453557 = score(doc=3036,freq=6.0), product of:
             0.0634953 = queryWeight, product of:
               1.2102818 = boost
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.014766776 = queryNorm
             0.54390746 = fieldWeight in 3036, product of:
               2.4494898 = tf(freq=6.0), with freq of:
                 6.0 = termFreq=6.0
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
         0.035760045 = weight(abstract_txt:relations in 3036) [ClassicSimilarity], result of:
           0.035760045 = score(doc=3036,freq=1.0), product of:
             0.10316108 = queryWeight, product of:
               1.2595866 = boost
               5.5462847 = idf(docFreq=468, maxDocs=44218)
               0.014766776 = queryNorm
             0.3466428 = fieldWeight in 3036, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               5.5462847 = idf(docFreq=468, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
         0.04336266 = weight(abstract_txt:language in 3036) [ClassicSimilarity], result of:
           0.04336266 = score(doc=3036,freq=2.0), product of:
             0.11730804 = queryWeight, product of:
               1.8995423 = boost
               4.1820874 = idf(docFreq=1834, maxDocs=44218)
               0.014766776 = queryNorm
             0.3696478 = fieldWeight in 3036, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               4.1820874 = idf(docFreq=1834, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
         0.108800724 = weight(abstract_txt:reasoning in 3036) [ClassicSimilarity], result of:
           0.108800724 = score(doc=3036,freq=1.0), product of:
             0.27290723 = queryWeight, product of:
               2.8972943 = boost
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.014766776 = queryNorm
             0.39867294 = fieldWeight in 3036, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
         0.23560694 = weight(abstract_txt:rules in 3036) [ClassicSimilarity], result of:
           0.23560694 = score(doc=3036,freq=5.0), product of:
             0.3219155 = queryWeight, product of:
               4.1626997 = boost
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.014766776 = queryNorm
             0.7318906 = fieldWeight in 3036, product of:
               2.236068 = tf(freq=5.0), with freq of:
                 5.0 = termFreq=5.0
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
         0.40708256 = weight(abstract_txt:rule in 3036) [ClassicSimilarity], result of:
           0.40708256 = score(doc=3036,freq=5.0), product of:
             0.44030735 = queryWeight, product of:
               4.5072193 = boost
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.014766776 = queryNorm
             0.92454183 = fieldWeight in 3036, product of:
               2.236068 = tf(freq=5.0), with freq of:
                 5.0 = termFreq=5.0
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.0625 = fieldNorm(doc=3036)
       0.24 = coord(6/25)
   

2. Monireh, E.; Sarker, M.K.; Bianchi, F.; Hitzler, P.; Doran, D.; Xie, N.: Reasoning over RDF knowledge bases using deep learning (2018) 0.18

   0.1846122 = sum of:
     0.1846122 = product of:
       0.7692175 = sum of:
         0.028198173 = weight(abstract_txt:knowledge in 4553) [ClassicSimilarity], result of:
           0.028198173 = score(doc=4553,freq=4.0), product of:
             0.0634953 = queryWeight, product of:
               1.2102818 = boost
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.014766776 = queryNorm
             0.4440986 = fieldWeight in 4553, product of:
               2.0 = tf(freq=4.0), with freq of:
                 4.0 = termFreq=4.0
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
         0.024041293 = weight(abstract_txt:over in 4553) [ClassicSimilarity], result of:
           0.024041293 = score(doc=4553,freq=1.0), product of:
             0.090626 = queryWeight, product of:
               1.445913 = boost
               4.244485 = idf(docFreq=1723, maxDocs=44218)
               0.014766776 = queryNorm
             0.2652803 = fieldWeight in 4553, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.244485 = idf(docFreq=1723, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
         0.10662698 = weight(abstract_txt:graphs in 4553) [ClassicSimilarity], result of:
           0.10662698 = score(doc=4553,freq=2.0), product of:
             0.1696232 = queryWeight, product of:
               1.6151495 = boost
               7.11192 = idf(docFreq=97, maxDocs=44218)
               0.014766776 = queryNorm
             0.62861085 = fieldWeight in 4553, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               7.11192 = idf(docFreq=97, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
         0.14321034 = weight(abstract_txt:scalability in 4553) [ClassicSimilarity], result of:
           0.14321034 = score(doc=4553,freq=2.0), product of:
             0.20648575 = queryWeight, product of:
               1.7820305 = boost
               7.84674 = idf(docFreq=46, maxDocs=44218)
               0.014766776 = queryNorm
             0.69356036 = fieldWeight in 4553, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               7.84674 = idf(docFreq=46, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
         0.18844838 = weight(abstract_txt:reasoning in 4553) [ClassicSimilarity], result of:
           0.18844838 = score(doc=4553,freq=3.0), product of:
             0.27290723 = queryWeight, product of:
               2.8972943 = boost
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.014766776 = queryNorm
             0.6905218 = fieldWeight in 4553, product of:
               1.7320508 = tf(freq=3.0), with freq of:
                 3.0 = termFreq=3.0
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
         0.27869228 = weight(abstract_txt:logical in 4553) [ClassicSimilarity], result of:
           0.27869228 = score(doc=4553,freq=2.0), product of:
             0.5109093 = queryWeight, product of:
               5.606244 = boost
               6.1714344 = idf(docFreq=250, maxDocs=44218)
               0.014766776 = queryNorm
             0.5454829 = fieldWeight in 4553, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               6.1714344 = idf(docFreq=250, maxDocs=44218)
               0.0625 = fieldNorm(doc=4553)
       0.24 = coord(6/25)
   

3. Allen, B.: Logical reasoning and retrieval performance (1993) 0.17

   0.17131698 = sum of:
     0.17131698 = product of:
       1.0707312 = sum of:
         0.031213615 = weight(abstract_txt:performance in 5093) [ClassicSimilarity], result of:
           0.031213615 = score(doc=5093,freq=1.0), product of:
             0.07190389 = queryWeight, product of:
               1.0515896 = boost
               4.63042 = idf(docFreq=1171, maxDocs=44218)
               0.014766776 = queryNorm
             0.43410188 = fieldWeight in 5093, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.63042 = idf(docFreq=1171, maxDocs=44218)
               0.09375 = fieldNorm(doc=5093)
         0.1219197 = weight(abstract_txt:ability in 5093) [ClassicSimilarity], result of:
           0.1219197 = score(doc=5093,freq=5.0), product of:
             0.104291506 = queryWeight, product of:
               1.266469 = boost
               5.5765896 = idf(docFreq=454, maxDocs=44218)
               0.014766776 = queryNorm
             1.169028 = fieldWeight in 5093, product of:
               2.236068 = tf(freq=5.0), with freq of:
                 5.0 = termFreq=5.0
               5.5765896 = idf(docFreq=454, maxDocs=44218)
               0.09375 = fieldNorm(doc=5093)
         0.3264022 = weight(abstract_txt:reasoning in 5093) [ClassicSimilarity], result of:
           0.3264022 = score(doc=5093,freq=4.0), product of:
             0.27290723 = queryWeight, product of:
               2.8972943 = boost
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.014766776 = queryNorm
             1.1960188 = fieldWeight in 5093, product of:
               2.0 = tf(freq=4.0), with freq of:
                 4.0 = termFreq=4.0
               6.378767 = idf(docFreq=203, maxDocs=44218)
               0.09375 = fieldNorm(doc=5093)
         0.59119564 = weight(abstract_txt:logical in 5093) [ClassicSimilarity], result of:
           0.59119564 = score(doc=5093,freq=4.0), product of:
             0.5109093 = queryWeight, product of:
               5.606244 = boost
               6.1714344 = idf(docFreq=250, maxDocs=44218)
               0.014766776 = queryNorm
             1.157144 = fieldWeight in 5093, product of:
               2.0 = tf(freq=4.0), with freq of:
                 4.0 = termFreq=4.0
               6.1714344 = idf(docFreq=250, maxDocs=44218)
               0.09375 = fieldNorm(doc=5093)
       0.16 = coord(4/25)
   

4. Kolluri, V.; Metzler, D.P.: Knowledge guided rule learning (1999) 0.15

   0.14558959 = sum of:
     0.14558959 = product of:
       0.6066233 = sum of:
         0.01495434 = weight(abstract_txt:knowledge in 6550) [ClassicSimilarity], result of:
           0.01495434 = score(doc=6550,freq=2.0), product of:
             0.0634953 = queryWeight, product of:
               1.2102818 = boost
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.014766776 = queryNorm
             0.23551884 = fieldWeight in 6550, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
         0.037021473 = weight(abstract_txt:mining in 6550) [ClassicSimilarity], result of:
           0.037021473 = score(doc=6550,freq=1.0), product of:
             0.12789261 = queryWeight, product of:
               1.4024676 = boost
               6.1754265 = idf(docFreq=249, maxDocs=44218)
               0.014766776 = queryNorm
             0.28947312 = fieldWeight in 6550, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               6.1754265 = idf(docFreq=249, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
         0.040318478 = weight(abstract_txt:over in 6550) [ClassicSimilarity], result of:
           0.040318478 = score(doc=6550,freq=5.0), product of:
             0.090626 = queryWeight, product of:
               1.445913 = boost
               4.244485 = idf(docFreq=1723, maxDocs=44218)
               0.014766776 = queryNorm
             0.44488862 = fieldWeight in 6550, product of:
               2.236068 = tf(freq=5.0), with freq of:
                 5.0 = termFreq=5.0
               4.244485 = idf(docFreq=1723, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
         0.049111653 = weight(abstract_txt:large in 6550) [ClassicSimilarity], result of:
           0.049111653 = score(doc=6550,freq=2.0), product of:
             0.16632955 = queryWeight, product of:
               2.52886 = boost
               4.454089 = idf(docFreq=1397, maxDocs=44218)
               0.014766776 = queryNorm
             0.29526716 = fieldWeight in 6550, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               4.454089 = idf(docFreq=1397, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
         0.07902497 = weight(abstract_txt:rules in 6550) [ClassicSimilarity], result of:
           0.07902497 = score(doc=6550,freq=1.0), product of:
             0.3219155 = queryWeight, product of:
               4.1626997 = boost
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.014766776 = queryNorm
             0.24548358 = fieldWeight in 6550, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
         0.3861924 = weight(abstract_txt:rule in 6550) [ClassicSimilarity], result of:
           0.3861924 = score(doc=6550,freq=8.0), product of:
             0.44030735 = queryWeight, product of:
               4.5072193 = boost
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.014766776 = queryNorm
             0.87709737 = fieldWeight in 6550, product of:
               2.828427 = tf(freq=8.0), with freq of:
                 8.0 = termFreq=8.0
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.046875 = fieldNorm(doc=6550)
       0.24 = coord(6/25)
   

5. Taniguchi, S.: ¬A system for analyzing cataloguing rules : a feasibility study (1996) 0.14

   0.14009605 = sum of:
     0.14009605 = product of:
       0.87560034 = sum of:
         0.036568787 = weight(abstract_txt:quality in 4198) [ClassicSimilarity], result of:
           0.036568787 = score(doc=4198,freq=3.0), product of:
             0.07260232 = queryWeight, product of:
               1.0566845 = boost
               4.6528544 = idf(docFreq=1145, maxDocs=44218)
               0.014766776 = queryNorm
             0.50368625 = fieldWeight in 4198, product of:
               1.7320508 = tf(freq=3.0), with freq of:
                 3.0 = termFreq=3.0
               4.6528544 = idf(docFreq=1145, maxDocs=44218)
               0.0625 = fieldNorm(doc=4198)
         0.014099087 = weight(abstract_txt:knowledge in 4198) [ClassicSimilarity], result of:
           0.014099087 = score(doc=4198,freq=1.0), product of:
             0.0634953 = queryWeight, product of:
               1.2102818 = boost
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.014766776 = queryNorm
             0.2220493 = fieldWeight in 4198, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.5527887 = idf(docFreq=3442, maxDocs=44218)
               0.0625 = fieldNorm(doc=4198)
         0.27877387 = weight(abstract_txt:rules in 4198) [ClassicSimilarity], result of:
           0.27877387 = score(doc=4198,freq=7.0), product of:
             0.3219155 = queryWeight, product of:
               4.1626997 = boost
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.014766776 = queryNorm
             0.8659846 = fieldWeight in 4198, product of:
               2.6457512 = tf(freq=7.0), with freq of:
                 7.0 = termFreq=7.0
               5.236983 = idf(docFreq=638, maxDocs=44218)
               0.0625 = fieldNorm(doc=4198)
         0.54615855 = weight(abstract_txt:rule in 4198) [ClassicSimilarity], result of:
           0.54615855 = score(doc=4198,freq=9.0), product of:
             0.44030735 = queryWeight, product of:
               4.5072193 = boost
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.014766776 = queryNorm
             1.240403 = fieldWeight in 4198, product of:
               3.0 = tf(freq=9.0), with freq of:
                 9.0 = termFreq=9.0
               6.615483 = idf(docFreq=160, maxDocs=44218)
               0.0625 = fieldNorm(doc=4198)
       0.16 = coord(4/25)