File size: 42,888 Bytes
6fa4bc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
{
    "paper_id": "W11-0149",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T05:39:05.891821Z"
    },
    "title": "Semantic Relatedness from Automatically Generated Semantic Networks",
    "authors": [
        {
            "first": "Pia-Ramona",
            "middle": [],
            "last": "Wojtinnek",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Oxford University Computing Laboratory",
                "location": {}
            },
            "email": "[email protected]"
        },
        {
            "first": "Stephen",
            "middle": [],
            "last": "Pulman",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Oxford University Computing Laboratory",
                "location": {}
            },
            "email": "[email protected]"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "We introduce a novel approach to measuring semantic relatedness of terms based on an automatically generated, large-scale semantic network. We present promising first results that indicate potential competitiveness with approaches based on manually created resources.",
    "pdf_parse": {
        "paper_id": "W11-0149",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "We introduce a novel approach to measuring semantic relatedness of terms based on an automatically generated, large-scale semantic network. We present promising first results that indicate potential competitiveness with approaches based on manually created resources.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The quantification of semantic similarity and relatedness of terms is an important problem of lexical semantics. Its applications include word sense disambiguation, text summarization and information retrieval (Budanitsky and Hirst, 2006) . Most approaches to measuring semantic relatedness fall into one of two categories. They either look at distributional properties based on corpora (Finkelstein et al., 2002; Agirre et al., 2009) or make use of pre-existing knowledge resources such as WordNet or Roget's Thesaurus (Hughes and Ramage, 2007; Jarmasz, 2003) . The latter approaches achieve good results, but they are inherently restricted in coverage and domain adaptation due to their reliance on costly manual acquisition of the resource. In addition, those methods that are based on hierarchical, taxonomically structured resources are generally better suited for measuring semantic similarity than relatedness (Budanitsky and Hirst, 2006) . In this paper, we introduce a novel technique that measures semantic relatedness based on an automatically generated semantic network. Terms are compared by the similarity of their contexts in the semantic network. We present our promising initial results of this work in progress, which indicate the potential to compete with resource-based approaches while performing well on both, semantic similarity and relatedness.",
                "cite_spans": [
                    {
                        "start": 210,
                        "end": 238,
                        "text": "(Budanitsky and Hirst, 2006)",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 387,
                        "end": 413,
                        "text": "(Finkelstein et al., 2002;",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 414,
                        "end": 434,
                        "text": "Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 520,
                        "end": 545,
                        "text": "(Hughes and Ramage, 2007;",
                        "ref_id": "BIBREF8"
                    },
                    {
                        "start": 546,
                        "end": 560,
                        "text": "Jarmasz, 2003)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 917,
                        "end": 945,
                        "text": "(Budanitsky and Hirst, 2006)",
                        "ref_id": "BIBREF2"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "In our approach to measuring semantic relatedness, we first automatically build a large semantic network from text and then measure the similarity of two terms by the similarity of the local networks around their corresponding nodes. The semantic network serves as a structured representation of the occurring concepts, relations and attributes in the text. It is built by translating every sentence in the text into a network fragment based on semantic analysis and then merging these networks into a large network by mapping all occurrences of the same term into one node. Figure 1 (a) contains a sample text snippet and the network derived from it. In this way, concepts are connected across sentences and documents, resulting in a high-level view of the information contained.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 575,
                        "end": 583,
                        "text": "Figure 1",
                        "ref_id": "FIGREF0"
                    }
                ],
                "eq_spans": [],
                "section": "Similarity and Relatedness from semantic networks",
                "sec_num": "2"
            },
            {
                "text": "Our underlying assumption for measuring semantic relatedness is that semantically related nodes are connected to a similar set of nodes. In other words, we consider the context of a node in the network as a representation of its meaning. In contrast to standard approaches which look only at a type of context directly found in the text, e.g. words that occur within a certain window from the target word, our network-based context takes into account indirect connections between concepts. For example, in the text underlying the network in Fig. 2 , dissertation and module rarely co-occurred in a sentence, but the network shows a strong connection over student as well as over credit and work.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 541,
                        "end": 547,
                        "text": "Fig. 2",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Similarity and Relatedness from semantic networks",
                "sec_num": "2"
            },
            {
                "text": "We build the network incrementally by parsing every sentence, translating it into a small network fragment and then mapping that fragment onto the main network generated from all previous sentences. Our translation of sentences from text to network is based on the one used in the ASKNet system (Harrington and Clark, 2007) . It makes use of two NLP tools, the Clark and Curran parser (Clark and Curran, 2004) and the semantic analysis tool Boxer (Bos et al., 2004) , both of which are part of the C&C Toolkit 1 . The parser is based on Combinatory Categorial Grammar (CCG) and has been trained on 40,000 manually annotated sentences of the WSJ. It is both robust and efficient. Boxer is designed to convert the CCG parsed text into a logical representation based on Discourse Representation Theory (DRT). This intermediate logical form representation presents an abstraction from syntactic details to semantic core information. For example, the syntactical forms progress of student and student's progress have the same Boxer representation as well as the student who attends the lecture and the student attending the lecture. In addition, Boxer provides some elementary co-reference resolution.",
                "cite_spans": [
                    {
                        "start": 295,
                        "end": 323,
                        "text": "(Harrington and Clark, 2007)",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 385,
                        "end": 409,
                        "text": "(Clark and Curran, 2004)",
                        "ref_id": "BIBREF3"
                    },
                    {
                        "start": 447,
                        "end": 465,
                        "text": "(Bos et al., 2004)",
                        "ref_id": "BIBREF1"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Network Structure",
                "sec_num": "2.1"
            },
            {
                "text": "The translation from the Boxer output into a network is straightforward and an example is given in Figure 1 (b). The network structure distinguishes between object nodes (rectangular), relational nodes (diamonds) and attributes (rounded rectangles) and different types of links such as subject or object links.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 99,
                        "end": 107,
                        "text": "Figure 1",
                        "ref_id": "FIGREF0"
                    }
                ],
                "eq_spans": [],
                "section": "The Network Structure",
                "sec_num": "2.1"
            },
            {
                "text": "Students select modules from the published list and write a dissertation. Modules usually provide 15 credits each, but 30 credits are awarded for the dissertation. The student must discuss the topic of the final dissertation with their appointed tutor. The large unified network is then built by merging every occurrence of a concept (e.g. object node) into one node, thus accumulating the information on this concept. In the second example ( Figure ??) , the lecture node would be merged with occurrences of lecture in other sentences. Figure 2 gives a subset of a network generated from a few paragraphs taken from Oxford Student Handbooks. Multiple occurrences of the same relation between two object nodes are drawn as overlapping.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 443,
                        "end": 453,
                        "text": "Figure ??)",
                        "ref_id": null
                    },
                    {
                        "start": 537,
                        "end": 545,
                        "text": "Figure 2",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "The Network Structure",
                "sec_num": "2.1"
            },
            {
                "text": "We measure the semantic relatedness of two concepts by measuring the similarity of the surroundings of their corresponding nodes in the network. Semantically related terms are then expected to be connected to a similar set of nodes. We retrieve the network context of a specific node and determine the level of significance of each node in the context using spreading activation 2 . The target node is given an initial activation of a x = 10 * numberOfLinks(x) and is fired so that the activation spreads over its outand ingoing links to the surrounding nodes. They in turn fire if their received activation level exceeds a certain threshold. The activation attenuates by a constant factor in every step and a stable state is reached when no node in the network can fire anymore. In this way, the context nodes receive different levels of activation reflecting their significance.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "We derive a vector representation v(x) of the network context of x including only object nodes and their activation levels. The entries are",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "v i (x) = act x,ax (n i ) n i \u2208 {n \u2208 nodes | type(n) = object node}",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "The semantic relatedness of two target words is then measured by the cosine similarity of their context vectors.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "sim rel(x, y) = cos( v(x), v(y)) = v(x) \u2022 v(y) v(x) v(y)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "As spreading activation takes several factors into account, such as number of paths, length of paths, level of density and number of connections, this method leverages the full interconnected structure of the network.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Vector Space Model",
                "sec_num": "2.2"
            },
            {
                "text": "We evaluate our approach on the WordSimilarity-353 (Finkelstein et al., 2002) test collection, which is a commonly used gold standard for the semantic relatedness task. It provides average human judgments scores of the degree of relatedness for 353 word pairs. The collection contains classically similar word Approach Spearman (Strube and Ponzetto, 2006) Wikipedia 0.19-0.48 (Jarmasz, 2003) Roget's 0.55 (Hughes and Ramage, 2007) WordNet 0.55 (Agirre et al., 2009) WordNet 0.56 (Finkelstein et al., 2002) Web corpus, LSA 0.56 (Harrington, 2010) Sem. Network 0.62 (Agirre et al., 2009) WordNet+gloss 0.66 (Agirre et al., 2009) Web corpus 0.66 (Gabrilovich and Markovitch, 2007) pairs such as street -avenue and topically related pairs such as hotel -reservation. However, no distinction was made while judging and the instruction was to rate the general degree of semantic relatedness.",
                "cite_spans": [
                    {
                        "start": 51,
                        "end": 77,
                        "text": "(Finkelstein et al., 2002)",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 328,
                        "end": 355,
                        "text": "(Strube and Ponzetto, 2006)",
                        "ref_id": "BIBREF10"
                    },
                    {
                        "start": 376,
                        "end": 391,
                        "text": "(Jarmasz, 2003)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 405,
                        "end": 430,
                        "text": "(Hughes and Ramage, 2007)",
                        "ref_id": "BIBREF8"
                    },
                    {
                        "start": 444,
                        "end": 465,
                        "text": "(Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 479,
                        "end": 505,
                        "text": "(Finkelstein et al., 2002)",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 527,
                        "end": 545,
                        "text": "(Harrington, 2010)",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 564,
                        "end": 585,
                        "text": "(Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 605,
                        "end": 626,
                        "text": "(Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 643,
                        "end": 677,
                        "text": "(Gabrilovich and Markovitch, 2007)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "As a corpus we chose the British National Corpus (BNC) 3 . It is one of the largest standardized English corpora and contains approximately 5.9 million sentences. Choosing this text collection enables us to build a general purpose network that is not specifically created for the considered work pairs and ensures a realistic overall connectedness of the network as well as a broad coverage. In this paper we created a network from 2 million sentences of the BNC. It contains 27.5 million nodes out of which 635.000 are object nodes and the rest are relation and attribute nodes. The building time including parsing was approximately 4 days.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "Following the common practice in related work, we compared our scores to the human judgements using the Spearman rank-order correlation coefficient. The results can be found in Table 1 (a) with a comparison to previous results on the WordSimilarity-353 collection.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 177,
                        "end": 184,
                        "text": "Table 1",
                        "ref_id": "TABREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "Our first result over all word pairs is relatively low compared to the currently best performing systems. However, we noticed that many poorly rated word pairs contained at least one word with low frequency. Excluding these considerably improved the result to 0.50. On this reduced set of word pairs our scores are in the region of approaches which make use of the Wikipedia category network, the Word-Net taxonomic relations or Roget's thesaurus. This is a promising result as it indicates that our approach based on automatically generated networks has the potential of competing with those using manually created resources if we increase the corpus size.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "While our results are not competitive with the best corpus based methods, we can note that our current corpus is an order of magnitude smaller -2 million sentences versus 1 million full Wikipedia articles (Gabrilovich and Markovitch, 2007) or 215MB versus 1.6 Terabyte (Agirre et al., 2009) . The extent to which corpus size influences our results is subject to further research.",
                "cite_spans": [
                    {
                        "start": 205,
                        "end": 239,
                        "text": "(Gabrilovich and Markovitch, 2007)",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 269,
                        "end": 290,
                        "text": "(Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "We also evaluated our scores separately on the semantically similar versus the semantically related subsets of WordSim-353 following Agirre et al. (2009) (Table 1(b) ). Taking the same low-frequency cut as above, we can see that our approach performs equally well on both sets. This is remarkable as different methods tend to be more appropriate to calculate either one or the other (Agirre et al., 2009) . In particular, WordNet based measures are well known to be better suited to measure similarity than relatedness due to its hierarchical, taxonomic structure (Budanitsky and Hirst, 2006) . The fact that our system achieves equal results on the subset indicates that it matches human judgement of semantic relatedness beyond specific types of relations. This could be due to the associative structure of the network.",
                "cite_spans": [
                    {
                        "start": 133,
                        "end": 153,
                        "text": "Agirre et al. (2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 383,
                        "end": 404,
                        "text": "(Agirre et al., 2009)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 564,
                        "end": 592,
                        "text": "(Budanitsky and Hirst, 2006)",
                        "ref_id": "BIBREF2"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 154,
                        "end": 165,
                        "text": "(Table 1(b)",
                        "ref_id": "FIGREF0"
                    }
                ],
                "eq_spans": [],
                "section": "Evaluation",
                "sec_num": "3"
            },
            {
                "text": "Our approach is closely related to Harrington (2010) as our networks are built in a similar fashion and we also use spreading activation to measure semantic relatedness. In their approach, semantic relatedness of two terms a and b is measured by the activation b receives when a is fired. The core difference of this measurement to ours is that it is path-based while ours is context based. In addition, the corpus used was retrieved specifically for the word pairs in question while ours is a general-purpose corpus.",
                "cite_spans": [
                    {
                        "start": 35,
                        "end": 52,
                        "text": "Harrington (2010)",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "4"
            },
            {
                "text": "In addition, our approach is related to work that uses personalized PageRank or Random Walks on WordNet (Agirre et al., 2009; Hughes and Ramage, 2007) . Similar the spreading activation method presented here, personalized PageRank and Random Walks are used to provide a relevance distribution of nodes surrounding the target word to its meaning. In contrast to the approaches based on resources, our network is automatically built and therefore does not rely on costly, manual creation. In addition, compared to WordNet based measures, our method is potentially not biased towards relatedness due to similarity.",
                "cite_spans": [
                    {
                        "start": 104,
                        "end": 125,
                        "text": "(Agirre et al., 2009;",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 126,
                        "end": 150,
                        "text": "Hughes and Ramage, 2007)",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "4"
            },
            {
                "text": "We presented a novel approach to measuring semantic relatedness which first builds a large-scale semantic network and then determines the relatedness of nodes by the similarity of their surrounding local network. Our preliminary results of this ongoing work are promising and are in the region of several WordNet and Wikipedia link structure approaches. As future work, there are several ways of improvement we are going to investigate. Firstly, the results in Section 3 show the crucial influence of corpus size and occurrence frequency on the performance of our system. We will be experimenting with larger general networks (e.g. the whole BNC) as well as integration of retrieved documents for the low frequency terms. Secondly, the parameters and specific settings for the spreading activation algorithm need to be tuned. For example, the amount of initial activation of the target node determines the size of the context considered. Thirdly, we will investigate different vector representation variants. In particular, we can achieve a more fine-grained representation by also considering relation nodes in addition to object nodes. We believe that with these improvements our automatic semantic network approach will be able to compete with techniques based on manually created resources.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion and Outlook",
                "sec_num": "5"
            },
            {
                "text": "http://svn.ask.it.usyd.edu.au/trac/candc",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "The spreading activation algorithm is based onHarrington (2010)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "http://www.natcorp.ox.ac.uk/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "A study on similarity and relatedness using distributional and wordnet-based approaches",
                "authors": [
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Agirre",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Alfonseca",
                        "suffix": ""
                    },
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Hall",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Kravalova",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Pa\u015fca",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Soroa",
                        "suffix": ""
                    }
                ],
                "year": 2009,
                "venue": "NAACL '09",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Agirre, E., E. Alfonseca, K. Hall, J. Kravalova, M. Pa\u015fca, and A. Soroa (2009). A study on similarity and relatedness using distributional and wordnet-based approaches. In NAACL '09.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Wide-coverage semantic representations from a ccg parser",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Bos",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Clark",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Steedman",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [
                            "R"
                        ],
                        "last": "Curran",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Hockenmaier",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "COLING'04",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Bos, J., S. Clark, M. Steedman, J. R. Curran, and J. Hockenmaier (2004). Wide-coverage semantic representations from a ccg parser. In COLING'04.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Evaluating wordnet-based measures of lexical semantic relatedness",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Budanitsky",
                        "suffix": ""
                    },
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Hirst",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "Computational Linguistics",
                "volume": "32",
                "issue": "1",
                "pages": "13--47",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Budanitsky, A. and G. Hirst (2006). Evaluating wordnet-based measures of lexical semantic relatedness. Computational Linguistics 32(1), 13-47.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Parsing the wsj using ccg and log-linear models",
                "authors": [
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Clark",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [
                            "R"
                        ],
                        "last": "Curran",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "ACL'04",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Clark, S. and J. R. Curran (2004). Parsing the wsj using ccg and log-linear models. In ACL'04.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Placing search in context: the concept revisited",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Finkelstein",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Gabrilovich",
                        "suffix": ""
                    },
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Matias",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Rivlin",
                        "suffix": ""
                    },
                    {
                        "first": "Z",
                        "middle": [],
                        "last": "Solan",
                        "suffix": ""
                    },
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Wolfman",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Ruppin",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "ACM Trans. Inf. Syst",
                "volume": "20",
                "issue": "1",
                "pages": "116--131",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Finkelstein, L., E. Gabrilovich, Y. Matias, E. Rivlin, Z. Solan, G. Wolfman, and E. Ruppin (2002). Placing search in context: the concept revisited. ACM Trans. Inf. Syst. 20(1), 116-131.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Computing semantic relatedness using wikipedia-based explicit semantic analysis",
                "authors": [
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Gabrilovich",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Markovitch",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Gabrilovich, E. and S. Markovitch (2007). Computing semantic relatedness using wikipedia-based explicit semantic analysis. In IJCAI'07.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "A semantic network approach to measuring semantic relatedness",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Harrington",
                        "suffix": ""
                    }
                ],
                "year": 2010,
                "venue": "COLING'10",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Harrington, B. (2010). A semantic network approach to measuring semantic relatedness. In COLING'10.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Asknet: automated semantic knowledge network",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Harrington",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Clark",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "AAAI'07",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Harrington, B. and S. Clark (2007). Asknet: automated semantic knowledge network. In AAAI'07.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Lexical semantic relatedness with random graph walks",
                "authors": [
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Hughes",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Ramage",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "EMNLP-CoNLL'07",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hughes, T. and D. Ramage (2007). Lexical semantic relatedness with random graph walks. In EMNLP-CoNLL'07.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Roget's thesaurus as a lexical resource for natural language processsing",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Jarmasz",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jarmasz, M. (2003). Roget's thesaurus as a lexical resource for natural language processsing. Master's thesis, University of Ottawa.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Wikirelate! computing semantic relatedness using wikipedia",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Strube",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [
                            "P"
                        ],
                        "last": "Ponzetto",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "AAAI'06",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Strube, M. and S. P. Ponzetto (2006). Wikirelate! computing semantic relatedness using wikipedia. In AAAI'06.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "uris": null,
                "text": "(a) Sample text snippet and according network representation. (b) Example of translation from text to network over Boxer semantic analysis",
                "type_str": "figure",
                "num": null
            },
            "FIGREF1": {
                "uris": null,
                "text": "Subgraph displaying selected concepts and relations from sample network.",
                "type_str": "figure",
                "num": null
            },
            "TABREF1": {
                "type_str": "table",
                "html": null,
                "num": null,
                "text": "",
                "content": "<table><tr><td>: (a) Spearman ranking correlation coefficient results for our approach and comparison with</td></tr><tr><td>previous approaches. (b) Separate results for similarity and relatedness subset.</td></tr></table>"
            }
        }
    }
}