chunk_id
stringlengths
3
7
chunk
stringlengths
1
823
source_url
stringclasses
416 values
title
stringclasses
416 values
chunk_idx
int64
0
294
chunk_start_char
int64
0
139k
chunk_end_char
int64
303
139k
9_4
The dataset, which is available under a CC-BY 4.0 license in the CoNNL-2002 format, was developed for training an NER service for German legal documents in the EU project Lynx. Named Entity Recognition, NER, Legal Documents, Legal Domain, Corpus Creation, Corpus Annotation Introduction and Motivation Just like any ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
4
1,599
2,157
9_5
This is where content curation technologies based on text analytics come in rehm2016j. Such domain-specific semantic technologies enable the fast and efficient automated processing of heterogeneous document collections, extracting important information units and metadata such as, among others, named entities, numeric ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
5
2,157
2,680
9_6
Typically, NER is focused upon the identification of semantic categories such as person, location and organization but, especially in domain-specific applications, other typologies have been developed that correspond to task-, language- or domain-specific needs. With regard to the legal domain, the lack of freely avai...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
6
2,680
3,231
9_7
The work described in this paper was carried out under the umbrella of the project Lynx: Building the Legal Knowledge Graph for Smart Compliance Services in Multilingual Europe, a three-year EU-funded project that started in December 2017 BIBREF2. Its objective is the creation of a legal knowledge graph that contains...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
7
3,231
3,744
9_8
The project offers compliance-related services that are currently tested and validated in three use cases (UC): (i) UC1 aims to analyse contracts, enriching them with domain-specific semantic information (document structure, entities, temporal expressions, claims, summaries, etc.); (ii) UC2 focuses on compliance servi...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
8
3,744
4,407
9_9
The Lynx services are developed for several European languages including English, Spanish, and – relevant for this paper – German BIBREF6. Documents in the legal domain contain multiple references to named entities, especially domain-specific named entities, i. e., jurisdictions, legal institutions, etc. Legal docume...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
9
4,407
4,867
9_10
On the other hand, in concrete applications, crucial domain-specific entities need to be identified in a reliable way, such as designations of legal norms and references to other legal documents (laws, ordinances, regulations, decisions, etc.). However, most NER solutions operate in the general or news domain, which m...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
10
4,867
5,467
9_11
In this paper, we describe the development of a dataset of legal documents, which includes (i) named entities and (ii) temporal expressions. The remainder of this article is structured as follows. First, Section SECREF3 gives a brief overview of related work. Section SECREF4 describes, in detail, the rationale behind...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
11
5,467
6,042
9_12
Related Work Until now, NER has not received a lot of attention in the legal domain, developed approaches are fragmented and inconsistent with regard to their respective methods, datasets and typologies used. Among the related work, there is no agreement regarding the selection of relevant semantic categories from t...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
12
6,042
6,575
9_13
dozier2010named describe five classes for which taggers are developed based on dictionary lookup, pattern-based rules, and statistical models. These are jurisdiction (a geographic area with legal authority), court, title (of a document), doctype (category of a document), and judge. The taggers were tested with docume...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
13
6,575
7,078
9_14
On the NERC level, entities were divided in abstraction, act, document, organization, person, and non-entity. With regard to LKIF, company, corporation, contract, statute etc. are used. Unfortunately, the authors do not provide any details regarding the questions how the entities were categorised or if there is any co...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
14
7,078
7,670
9_15
References are identified based on the rules described by landthaler2016unveiling. The authors created an evaluation dataset of 20 court decisions. Annotation of the Dataset In the following, we describe the rationale behind the annotation of the dataset including the definition of the various semantic classes and t...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
15
7,670
8,235
9_16
The examples (UNKREF6–UNKREF8) show corresponding sentences that contain the named mention `John', the nominal mention `the boy' and the pronominal mention `he'. This distinction between names on the one hand and pronominal or nominal mentions on the other can also be applied to the broad semantic set of named entitie...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
16
8,235
8,726
9_17
The BGB regulates the legal relations between private persons. The law regulates the legal relations [...]. It regulates the legal relations [...]. Annotation of the Dataset ::: Named Entities vs. Legal Entities ::: Legal Entity Basically, legal entities are either designations or references. A designation (or na...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
17
8,726
9,204
9_18
The title of the Act on the Federal Constitutional Court is: `Gesetz über das Bundesverfassungsgericht (Bundesverfassungsgerichtsgesetz – BVerfGG)', where `Gesetz über das Bundesverfassungsgericht' is the long title, `Bundesverfassungsgerichtsgesetz' is the short title, and `BVerfGG' is the abbreviation. A reference t...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
18
9,204
9,727
9_19
Annotation of the Dataset ::: Named Entities vs. Legal Entities ::: Personal Data A fundamental characteristic of the published decisions, that are the basis of our dataset, is that all personal information have been anonymised for privacy reasons. This affects the classes person, location and organization. Dependin...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
19
9,727
10,350
9_20
that are part of this named entity (UNKREF16). Fernsehmoderator G. PER `television presenter G.' Firma X... UN `company X...' in der A-Straße STR in ... ST `in the A-Street in ...' Annotation of the Dataset ::: Semantic Classes We defined 19 fine-grained semantic classes. The (proto)typical classes are person,...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
20
10,350
10,774
9_21
These are the coarse-grained classes legal norm, case-by-case regulation, court decision and legal literature. The classes legal norm and case-by-case regulation include designations and references, while court decision and legal literature include only references. In the process of developing the typology and annota...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
21
10,774
11,440
9_22
Continent was integrated into landscape. The specification of the 19 fine-grained classes was motivated by the need for distinguishing entities in the legal domain. A first distinction was made between standards and binding acts. Standards, which belong to legal norm, are legal rules adopted by a legislative body in ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
22
11,440
12,039
9_23
It includes regulation (arrangements or instructions on subjects) and contract (agreements between subjects). In addition, court decision and legal literature, which are important in the decision making process, were put into their own categories. Within person, we distinguish between judge and lawyer, key roles ment...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
23
12,039
12,682
9_24
Annotation of the Dataset ::: Semantic Classes ::: Person The coarse-grained class person PER contains the fine-grained classes judge RR, lawyer AN and person PER (such as accused, plaintiff, defendant, witness, appraiser, expert, etc.), who are involved in a court process and mentioned in a decision. In example (UN...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
24
12,682
13,087
9_25
Zwar ist Paul Kirchhof RR mit dem Vizepräsidenten Kirchhof PER als dessen Bruder in der Seitenlinie im zweiten Grade verwandt... `Although Paul Kirchhof is related to the Vice President Kirchhof as his brother in the second-degree sidelines...' Annotation of the Dataset ::: Semantic Classes ::: Location The coarse...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
25
13,087
13,663
9_26
Street (UNKREF23) refers to avenues, squares, municipalities, attractions etc., i. e., named entities within a city or a village. Landscape (UNKREF24) includes continents, lakes, rivers and other geographical objects. ... hat bislang nur das Land Mecklenburg-Vor-ForestGreen!50I pommern LD Gebrauch gemacht. `So far, ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
26
13,663
14,129
9_27
Februar 2016 zugrunde ... `The arrest warrant is based on a decision of the Appeal Court in Bucharest of 18 February 2016 ...' Zwar legt der Bezug auf die Grenzwertüberschreitung 2015 insbesondere in der Cornelius-GreenYellowI straße STR ... `Admittedly, the reference to the exceedance of the 2015 threshold applies...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
27
14,129
14,542
9_28
dort angeboten werden ... `... come from the region around the river Main or are offered there...' Annotation of the Dataset ::: Semantic Classes ::: Organization The coarse-grained class organization ORG is divided into public/social, state and economic institutions. Social and public institutions such as parties,...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
28
14,542
15,025
9_29
Institution INN (UNKREF27) contain public administrations, including federal and state ministries and the constitutional bodies of the Federal Republic of Germany at the federal and state level, i. e., the Federal Government, the Federal Council, the Bundestag, the state parliaments and governments. Company UN (UNKREF...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
29
15,025
15,610
9_30
`The state government of Rhineland-Palatinate refrained from commenting.' ... eingeführte Smartphone-Modellreihe des US-amerikanischen Unternehmens Apple UN ... `... introduced smartphone model series of the US company Apple ...' Court designations play a central role in decisions, which is why they are collected ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
30
15,610
16,170
9_31
Furthermore, brands are often discussed in decisions of the Federal Patent Court. They are subsumed under brand MRK, which can be contextual and semantically ambiguous, such as `Becker' from (UNKREF30), which has evolved from a personal name.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
31
16,170
16,413
9_32
Diesen Anspruch hat das LSG Mecklenburg-RubineRed!50I Vorpommern GRT mit Urteil vom 22.2.2017 verneint ... `This claim was rejected by the LSG Mecklenburg-Vorpommern by judgment of 22.2.2017 ...' Vorliegend stehen sich die Widerspruchsmarke Becker Mining MRK und die angegriffene Marke Becker MRK gegenüber. `In the...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
32
16,413
16,830
9_33
Annotation of the Dataset ::: Semantic Classes ::: Legal Norms Norms are divided according to their legal status into the fine-grained classes of law GS, ordinance VO and European legal norm EUN. Law is composed of the standards adopted and designated by the legislature (Bundestag, Bundesrat, Landtag). Ordinance inc...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
33
16,830
17,371
9_34
Example (UNKREF32) includes a reference to the `Part-Time and Limited Term Employment Act' and the designation 'Basic Law'. The complex reference consists of the reference to the particular section of the law, its name and abbreviation (in brackets), date of issue, the reference in parenthesis and the details of the ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
34
17,371
17,933
9_35
... § 14 Absatz 2 Satz 2 des Gesetzes über Teil- RedOrange!70I zeitarbeit und IRedOrange!70Ibefristete Arbeitsverträge RedOrange!70I(Tz-RedOrange!70I BfG) vom 21.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
35
17,933
18,097
9_36
Dezember 2000 (Bundesgesetz- RedOrange!70I blatt Seite 1966), zuletzt geändert durch GesetzRedOrange!70I vomRedOrange!70IRedOrange!70I20.RedOrange!70IDezemberRedOrange!70I2011 (Bundesgesetzblatt IRedOrange!70I Seite 2854 ) RedOrange!70I GS, ist nach Maßgabe der Gründe mit dem Grundgesetz GS vereinbar.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
36
18,097
18,400
9_37
`... section 14 paragraph 2 sentence 2 of the Law on Part-Time and Limited Term Employment Act (TzBfG) of 21 December 2000 (Federal Law Gazette I, page 1966), as last amended by the Law of 20 December 2011 (Federal Law Gazette I, page 2854), shall be published in accordance with the reasons compatible with the Basic ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
37
18,400
18,797
9_38
35 para. 6 StVO...' Annotation of the Dataset ::: Semantic Classes ::: Case-by-case Regulation The class case-by-case regulation REG contains individual binding acts. These include regulation VS and contract VT. Regulation is an internal order or instruction from a superordinate authority to a subordinate, regulatin...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
38
18,797
19,401
9_39
Some designations and references from these classes are similar to legal norm (UNKREF35, UNKREF36). ... insbesondere durch die Richtlinien zur Be-Peach!70I wertung des Grundvermögens –BewRGr– vom Peach!70I 19. September 1966 (BStBl I, S. 890) VS. `... in particular by the Guidelines for the Valuation of Real Estate ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
39
19,401
19,772
9_40
... fand der Manteltarifvertrag für die Beschäf-Goldenrod!70I tigten der Mitglieder der TGAOK VT (BAT/Goldenrod!70I AOK-Neu VT) vom 7. August 2003 Anwendung. `... the Collective Agreement for the Employees of Members of TGAOK (BAT/AOK-New) was applied of 7 August 2003 ...' Annotation of the Dataset ::: Semantic Cla...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
40
19,772
20,178
9_41
It does not have any subclasses, the coarsed and fine-grained versions are identical. In court decision, the name of the official decision-making collection, the volume and the numbered article are cited. Often mentioned are also the court, if necessary the decision type, date and file number. Example (UNKREF39) cites...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
41
20,178
20,743
9_42
Annotation of the Dataset ::: Semantic Classes ::: Legal Literature Legal literature LIT also contains references, but they refer to legal commentaries, legislative material, legal textbooks and monographs. The commentary in example (UNKREF39) includes the details of an author's and/or publisher's name, the name of ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
42
20,743
21,356
9_43
... vgl zB BVerfGE 62, 1, 45 RS; BVerfGEDandelion!70I 119, 96, 179 RS; BSG SozR 4–2500 § 62 NrDandelion!70I 8 RdNr 20 f RS; Hauck/Wiegand, KrV 2016,Tan!60!Bittersweet!70!whiteI 1, 4 LIT ... `... cf. i.e.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
43
21,356
21,562
9_44
BVerfGE 62, 1, 45; BVerfGE 119, 96, 179; BSG SozR 4–2500 § 62 Nr 8 RdNr 20 f; Hauck/Wiegand, KrV 2016, 1, 4 ...' Description of the Dataset The dataset, which also includes annotation guidelines, is freely available under a CC-BY 4.0 license.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
44
21,562
21,807
9_45
The named entity annotations adhere to the CoNLL-2002 format BIBREF14, while time expressions were annotated using TimeML BIBREF15. Description of the Dataset ::: Original Source Documents Legal documents are a rather heterogeneous class, which also manifests in their linguistic properties, including the use of name...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
45
21,807
22,416
9_46
When comparing legal documents such as laws, court decisions or administrative regulations, decisions are the best option. In laws and administrative regulations, the frequencies of person, location and organization are not high enough for NER experiments. Court decisions, on the other hand, include person, location, ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
46
22,416
22,943
9_47
The documents originate from seven federal courts: Federal Labour Court (BAG), Federal Fiscal Court (BFH), Federal Court of Justice (BGH), Federal Patent Court (BPatG), Federal Social Court (BSG), Federal Constitutional Court (BVerfG) and Federal Administrative Court (BVerwG). From the table of contents, 107 document...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
47
22,943
23,308
9_48
The data was collected from the XML documents, i. e., it was extracted from the XML elements Mitwirkung, Titelzeile, Leitsatz, Tenor, Tatbestand, Entscheidungsgründe, Gründen, abweichende Meinung, and sonstiger Titel. The metadata at the beginning of the documents (name of court, date of decision, file number, Europea...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
48
23,308
23,770
9_49
The extracted data was split into sentences, tokenised using SoMaJo BIBREF16 and manually annotated in WebAnno BIBREF17. The annotated documents are available in CoNNL-2002. The information originally represented by and through the XML markup was lost in the conversion process. We decided to use CoNNL-2002 because ou...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
49
23,770
24,277
9_50
Nevertheless, it is possible, of course, to re-insert the annotated information back into the XML documents. Description of the Dataset ::: Annotation of Named Entities The dataset consists of 66,723 sentences with 2,157,048 tokens (incl. punctuation), see Table . The sizes of the seven court-specific datasets varie...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
50
24,277
24,741
9_51
19–23 %. The Federal Patent Court (BPatG) dataset contains the lowest number of annotated entities (10.41 %). The dataset includes two different versions of annotations, one with a set of 19 fine-grained semantic classes and another one with a set of 7 coarse-grained classes (Table ). There are 53,632 annotated entit...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
51
24,741
25,186
9_52
Overall, the most frequent entities are law GS (34.53 %) and court decision RS (23.46 %). The other legal classes (ordinance VO, European legal norm EUN, regulation VS, contract VT, and legal literature LIT) are much less frequent (1–6 % each). Even less frequent (less than 1 %) are lawyer AN, street STR, landscape LD...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
52
25,186
25,649
9_53
More than half of city and street, about 55 %, have also been modified. Landscape and organization are affected as well, with 40 % and 15 % of the occurrences edited accordingly. However, anonymisation is typically not applied to judge, country, institution and court (1–5 %). The dataset was originally annotated by t...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
53
25,649
26,257
9_54
For the sentence extraction we paid special attention to the anonymised mentions of person, location or organization entities, because these are usually explained at their first mention. The resulting sample consisted of 2005 sentences with a broad variety of different entities (3 % of all sentences from each federal ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
54
26,257
26,864
9_55
Differences were in the identification of court decision and legal literature. Some unusual references of court decision (consisting only of decision type, court, date, file number) were not annotated such as `Urteil des Landgerichts Darmstadt vom 16. April 2014 – 7 S 8/13 –'. Apart from missing legal literature annot...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
55
26,864
27,329
9_56
35', `Bekanntmachung des BMG gemäß §§ 295 und 301 SGB V zur Anwendung des OPS vom 21.10.2010'). The second annotator had difficulties annotating the class law, not all instances were identified (`§ 272 Abs. 1a und 1b HGB', `§ 3c Abs.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
56
27,329
27,564
9_57
2 Satz 1 EStG'), others only partially (`§ 716 in Verbindung mit' in `§ 716 in Verbindung mit §§ 321 , 711 ZPO'). Some titles of contract were not recognised and annotated (`BAT', `TV-L', `TVÜ-Länder' etc.). This evaluation has revealed deficiencies in the annotation guidelines, especially regarding court decision an...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
57
27,564
28,067
9_58
Description of the Dataset ::: Annotation of Time Expressions All court decisions were annotated automatically for time expressions using a customised version of HeidelTime BIBREF19, which was adapted to the legal domain BIBREF20. This version of Heideltime achieves an F$_1$ value of 89.1 for partial identification ...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
58
28,067
28,579
9_59
It also includes expressions such as `present', `former' or `future'. DURATION describes time periods such as `two hours' or `six years'. SET describes a set of times/periods (`every day', `twice a week'). TIME describes a time expression (`13:12', `tomorrow afternoon'). Expressions with a granularity less than 24 hou...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
59
28,579
29,076
9_60
94 of which are of type DATE. ...vgl. BGH, Beschluss vom blue$<$TIMEX3 redtid=Purple"t14" redtype=Purple"DATE" redvalue=Purple”1999-02-03”blue$>$3. Februar 1999blue$<$/TIMEX3$>$ – 5 StR 705/98, juris Rn. 2 ... Evaluation The dataset was thoroughly evaluated, see leitner2019fine for more details.
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
60
29,076
29,376
9_61
As state of the art models, Conditional Random Fields (CRFs) and bidirectional Long-Short Term Memory Networks (BiLSTMs) were tested with the two variants of annotation. For CRFs, these are: CRF-F (with features), CRF-FG (with features and gazetteers), CRF-FGL (with features, gazetteers and lookup).
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
61
29,376
29,677
9_62
For BiLSTM, we used models with pre-trained word embeddings BIBREF22: BiLSTM-CRF BIBREF23, BiLSTM-CRF+ with character embeddings from BiLSTM BIBREF24, and BiLSTM-CNN-CRF with character embeddings from CNN BIBREF25. To evaluate the performance we used stratified 10-fold cross-validation. As expected, BiLSTMs perform be...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
62
29,677
30,117
9_63
CRFs reach up to 93.23 F$_1$ for the fine-grained classes and 93.22 F$_1$ for the coarse-grained ones. Both models perform best for judge, court and law. Conclusions and Future Work We describe a dataset that consists of German legal documents. For the annotation, we specified a typology of characteristic semantic c...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
63
30,117
30,597
9_64
A functional service based on the work described in this paper will be made available through the European Language Grid BIBREF26. In terms of future work, we will look into approaches for extending and further optimizing the dataset. We will also perform additional experiments with more recent state of the art appro...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
64
30,597
31,200
9_65
We also plan to produce an XML version of the dataset that also includes the original XML annotations. Acknowledgements This work has been partially funded by the project Lynx, which has received funding from the EU's Horizon 2020 research and innovation programme under grant agreement no. 780602, see http://www.lyn...
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
65
31,200
31,533
9_66
Table 1: Dataset size (tokens, sentences, annotated tokens) Table 2: Distribution of fine-grained (f) and coarse-grained (c) classes in the dataset Table 3: Distribution of time expressions in the dataset Table 4: Precision, recall and F1 values of the CRF and BiLSTM models for the fine- and coarse-grained classes
https://arxiv.org/abs/2003.13016
A Dataset of German Legal Documents for Named Entity Recognition
66
31,533
31,853
10_0
Character-Level Models versus Morphology in Semantic Role Labeling Character-level models have become a popular approach specially for their accessibility and ability to handle unseen data. However, little is known on their ability to reveal the underlying morphological structure of a word, which is a crucial skill fo...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
0
0
601
10_1
We conduct an in-depth error analysis for each morphological typology and analyze the strengths and limitations of character-level models that relate to out-of-domain data, training data size, long range dependencies and model complexity. Our exhaustive analyses shed light on important characteristics of character-lev...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
1
601
1,307
10_2
This assumption has two major shortcomings especially for languages with rich morphology: (1) inability to handle unseen or out-of-vocabulary (OOV) word-forms (2) inability to exploit the regularities among word parts. The limitations of word embeddings are particularly pronounced in sentence-level semantic tasks, esp...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
2
1,307
1,844
10_3
Here the stems köy (village) and sendika (union) function similarly in semantic terms with respect to the verb come (as the origin of the agents of the verb), where şehir (town) and meclis (council) both function as the end point. These semantic similarities are determined by the common word parts shown in bold. Howev...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
3
1,844
2,337
10_4
Therefore, for a successful semantic application, the model should be able to capture both the regularities, i.e, morphological tags and the irregularities, i.e, lemmas of the word. Morphological analysis already provides the aforementioned information about the words. However access to useful morphological features ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
4
2,337
2,969
10_5
However the extent to which these tasks depend on morphology is small; and their relation to semantics is weak. Hence, little is known on their true ability to reveal the underlying morphological structure of a word and their semantic capabilities. Furthermore, their behaviour across languages from different families;...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
5
2,969
3,531
10_6
To achieve this, we perform a case study on semantic role labeling (SRL), a sentence-level semantic analysis task that aims to identify predicate-argument structures and assign meaningful labels to them as follows: $[$ Villagers $]$ comers came $[$ to town $]$ end point We use a simple method based on bidirectiona...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
6
3,531
4,008
10_7
The gold morphology serves as the upper bound for us to compare and analyze the performances of character-level models on languages of varying morphological typologies. We carry out an exhaustive error analysis for each language type and analyze the strengths and limitations of character-level models compared to morph...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
7
4,008
4,751
10_8
Our experiments and analysis reveal insights such as: Method Formally, we generate a label sequence $\vec{l}$ for each sentence and predicate pair: $(s,p)$ . Each $l_t\in \vec{l}$ is chosen from $\mathcal {L}=\lbrace \mathit {roles \cup nonrole}\rbrace $ , where $roles$ are language-specific semantic roles (mostly ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
8
4,751
5,164
10_9
Given $\theta $ as model parameters and $g_t$ as gold label for $t_{th}$ token, we find the parameters that minimize the negative log likelihood of the sequence: $$\hat{\theta }=\underset{\theta }{\arg \min } \left( -\sum _{t=1}^n log (p(g_t|\theta ,s,p)) \right)$$ (Eq. 7) Label probabilities, $p(l_t|\theta ,s,p...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
9
5,164
5,531
10_10
First, the word encoding layer splits tokens into subwords via $\rho $ function. $$\rho (w) = {s_0,s_1,..,s_n}$$ (Eq. 8) As proposed by BIBREF0 , we treat words as a sequence of subword units. Then, the sequence is fed to a simple bi-LSTM network BIBREF15 , BIBREF16 and hidden states from each direction are weig...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
10
5,531
5,996
10_11
$$hs_f, hs_b = \text{bi-LSTM}({s_0,s_1,..,s_n}) \\ \vec{w} = W_f \cdot hs_f + W_b \cdot hs_b + b$$ (Eq. 9) There may be more than one predicate in the sentence so it is crucial to inform the network of which arguments we aim to label. In order to mark the predicate of interest, we concatenate a predicate flag $p...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
11
5,996
6,350
10_12
$$\vec{x_{t}} = [\vec{w};pf_t]$$ (Eq. 10) Final vector, $\vec{x_t}$ serves as an input to another bi-LSTM unit. $$\vec{h_{f}, h_{b}} = \text{bi-LSTM}(x_{t})$$ (Eq. 11) Finally, the label distribution is calculated via softmax function over the concatenated hidden states from both directions.
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
12
6,350
6,655
10_13
$$\vec{p(l_t|s,p)} = softmax(W_{l}\cdot [\vec{h_{f}};\vec{h_{b}}]+\vec{b_{l}})$$ (Eq. 12) For simplicity, we assign the label with the highest probability to the input token. . Subword Units We use three types of units: (1) words (2) characters and character sequences and (3) outputs of morphological analysis....
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
13
6,655
7,064
10_14
Table 1 shows sample outputs of various $\rho $ functions. Here, char function simply splits the token into its characters. Similar to n-gram language models, char3 slides a character window of width $n=3$ over the token. Finally, gold morphological features are used as outputs of morph-language. Throughout this pape...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
14
7,064
7,621
10_15
As an exception, it outputs additional information for some languages, such as parts-of-speech tags for Turkish. Word segmenters such as Morfessor and Byte Pair Encoding (BPE) are other commonly used subword units. Due to low scores obtained from our preliminary experiments and unsatisfactory results from previous stu...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
15
7,621
7,981
10_16
Experiments We use the datasets distributed by LDC for Catalan (CAT), Spanish (SPA), German (DEU), Czech (CZE) and English (ENG) BIBREF17 , BIBREF18 ; and datasets made available by BIBREF19 , BIBREF20 for Finnish (FIN) and Turkish (TUR) respectively . Datasets are provided with syntactic dependency annotations and ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
16
7,981
8,540
10_17
Here, #pred is number of predicates, and #role refers to number distinct semantic roles that occur more than 10 times. More detailed statistics about the datasets can be found in BIBREF27 , BIBREF19 , BIBREF20 . Experimental Setup To fit the requirements of the SRL task and of our model, we performed the following: ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
17
8,540
9,047
10_18
For the sake of memory efficiency and performance, we used an abbreviation (e.g., CFdT) for each MWE during training and testing. Original dataset defines its own format of semantic annotation, such as 17:PBArgM_mod $\mid $ 19:PBArgM_mod meaning the node is an argument of $17_{th}$ and $19_{th}$ tokens with ArgM-mod ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
18
9,047
9,522
10_19
Words are splitted from derivational boundaries in the original dataset, where each inflectional group is represented as a separate token. We first merge boundaries of the same word, i.e, tokens of the word, then we use our own $\rho $ function to split words into subwords. We lowercase all tokens beforehand and pla...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
19
9,522
10,103
10_20
We used gradient clipping and early stopping to prevent overfitting. Stochastic gradient descent is used as the optimizer. The initial learning rate is set to 1 and reduced by half if scores on development set do not improve after 3 epochs. We use the provided splits and evaluate the results with the official evaluati...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
20
10,103
10,653
10_21
Most of the early SRL work report combined scores (argument labeling with predicate sense disambiguation (PSD)). However, PSD is considered a simpler task with higher F1 scores . Therefore, we believe omitting PSD helps us gain more useful insights on character level models. Results and Analysis Our main results on ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
21
10,653
11,260
10_22
IOW and IOC values are calculated on the test set. The biggest improvement over the word baseline is achieved by the models that have access to morphology for all languages (except for English) as expected. Character trigrams consistently outperformed characters by a small margin. Same pattern is observed on the resu...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
22
11,260
11,746
10_23
We analyze the results separately for agglutinative and fusional languages and reveal the links between certain linguistic phenomena and the IOC, IOW values. Similarity between models One way to infer similarity is to measure diversity. Consider a set of baseline models that are not diverse, i.e., making similar err...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
23
11,746
12,378
10_24
Suppose that a prediction $p_{i}$ is generated for each token by a model $m_i$ , $i \in n$ , then the final prediction is calculated from these predictions by: $$p_{final} = f(p_0, p_1,..,p_n|\phi )$$ (Eq. 36) where $f$ is the combining function with parameter $\phi $ . The simplest global approach is averaging ...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
24
12,378
12,781
10_25
Mean function combines model outputs linearly, therefore ignores the nonlinear relation between base models/units. In order to exploit nonlinear connections, we learn the parameters $\phi $ of $f$ via a simple linear layer followed by sigmoid activation. In other words, we train a new model that learns how to best com...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
25
12,781
13,393
10_26
To ensure that the only factor contributing to the diversity of the learners is the input representation, all parameters, training data and model settings are left unchanged. Our results are given in Table 4 . IOB shows the improvement over the best of the baseline models in the ensemble. Averaging and stacking metho...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
26
13,393
14,014
10_27
Amongst the first set, we observe that the improvement gained by character-morphology ensembles is higher (shown with green) than ensembles between characters and character trigrams (shown with red), whereas the opposite is true for the second set of languages. It can be interpreted as character level models being mor...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
27
14,014
14,502
10_28
Limitations and Strengths To expand our understanding and reveal the limitations and strengths of the models, we analyze their ability to handle long range dependencies, their relation with training data and model size; and measure their performances on out of domain data. Long Range Dependencies Long range depend...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
28
14,502
15,094
10_29
The unit of measure is number of tokens between the two; and argument is defined as the head of the argument phrase in accordance with dependency-based SRL task. For that purpose, we created bins of [0-4], [5-9], [10-14] and [15-19] distances. Then, we have calculate F1 scores for arguments in each bin. Due to low num...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
29
15,094
15,581
10_30
We observe that either char or char3 closely follows the oracle for all languages. The gap between the two does not increase with the distance, suggesting that the performance gap is not related to long range dependencies. In other words, both characters and the oracle handle long range dependencies equally well. Tra...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
30
15,581
16,210
10_31
4 . Apparently as the data size increases, the performances of both models logarithmically increase - with a varying speed. To speak in statistical terms, we fit a logarithmic curve to the observed F1 scores (shown with transparent lines) and check the x coefficients, where x refers to the number of sentences. This co...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
31
16,210
16,699
10_32
It can be interpreted as: in the presence of more training data, char3 may surpass the oracle; i.e., char3 relies on data more than the oracle. Out-of-Domain (OOD) Data As part of the CoNLL09 shared task BIBREF27 , out of domain test sets are provided for three languages: Czech, German and English. We test our model...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
32
16,699
17,202
10_33
The dramatic drop in German oracle model is due to the high lemma OOV rate which is a consequence of keeping compounds as a single lemma. Czech oracle model performs reasonably however is unable to beat the generalization power of the char3 model. Furthermore, the scores of the character models in Table 5 are higher t...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
33
17,202
17,801
10_34
Model Size Throughout this paper, our aim was to gain insights on how models perform on different languages rather than scoring the highest F1. For this reason, we used a model that can be considered small when compared to recent neural SRL models and avoided parameter search. However, we wonder how the models behav...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
34
17,801
18,450
10_35
This indicates that morphological signals help to extract more complex linguistic features that have semantic clues. Predicted Morphological Tags Although models with access to gold morphological tags achieve better F1 scores than character models, they can be less useful a in real-life scenario since they require g...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
35
18,450
19,056
10_36
5 , show that (except for Czech), predicted morphological tags are not as useful as characters alone. Conclusion Character-level neural models are becoming the defacto standard for NLP problems due to their accessibility and ability to handle unseen data. In this work, we investigated how they compare to models with...
https://arxiv.org/abs/1805.11937
Character-Level Models versus Morphology in Semantic Role Labeling
36
19,056
19,556