Identification of entities in the Linked Data collection “Rainis and Aspazija” (RunA)

This paper offers an insight into the work in progress — development of the Linked Data collection RunA http://runa.lnb.lv/ on two Latvian poets and opinion leaders (Rainis — J ā nis Pliekš ā ns, 1865-1929 and Aspazija, Johanna Em ī lija Lizete Rozenberga, 1865-1943) at the National Library of Latvia (NLL). Authors report findings from the theoretical study on identification of RunA collection entities and reality. As collection’s textual materials comprise things and objects of real life, drawbacks in comprehension of NLL authority data are evaluated and perspective with URI identification is emphasised. The authors also address the need to deal with persistent identification for abstract entities. Considering that RunA collection comprises a lot of textual materials (unstructured data) which are to be mutually linked by annotating and identifying key entities (works, persons, institutions, concepts, places, events, etc.), the RDA rules from RunA collection point of view are examined. The main challenges which the creators of RunA faced during pilot project are presented along with corresponding solutions. Conclusions on possible RDA development influence on named entity identification in RunA are reported.

Rainis and Aspazija's heritage is placed in libraries, archives, museums and private collections inside and outside Latvia (Switzerland, Lithuania, Russia).The initial collection (approximately 500 objects) provides an opportunity to navigate through works of both poets (34 of Aspazija and 49 of Rainis) linking them with other textual, visual and audiovisual objects of the collection: small part of their correspondence, archival documents, photos, posters, audio and video files.
RunA discloses the diversity of both poets' heritage.It offers possibilities that have not been used in Latvia before to discover and to expose values from memory institutions and private repositories as an interlinked web collection.Users of this collection can navigate from specific objects to related historical events (timeline of poets' works and related events) as well as to external objects: related persons, institutions, places, works, etc. (visualization of links between objects; object link graph, i.e. entity "relationship tree").
The pilot collection was created with three main aims in mind.The first is to develop a novel resource for digital humanities to include as many types of digital objects as possible.The second one is to test in practice co-operation capabilities with other memory institutions, such as the National Archives of Latvia, the Institute of Literature, Folklore and Art of the University of Latvia, the Association of Memorial Museums and the Literature and Music Museum.The third one is to publish corresponding NLL bibliographic, authority and other metadata on the Web as Linked Data.
Besides that, RunA is an experimental collection for putting the following new approaches into practice: evaluating new library and web standards; re-using bibliographic and authority data in Web environment, linking them with entities from other present metadata sets and textual documents (literary work annotations and annotated transcripts of correspondence including scientific comments).These textual materials contain a big amount of named entities like people, places, events, etc. that are referenced in bibliographic and authority data.This is an attempt to extract these entities with their attributes and relations from MARC data and link them with marked entities from texts and other type of metadata in Dublin Core (DC).

Findings from the theoretical study on identification of RunA pilot collection entities
The above mentioned co-operating institutions contributed specific kinds of materials they possessmetadata (authority and bibliographic records), digital copies of letters from Rainis and Aspazija, short descriptions of their creative works (digital texts), legal documents and documents related to education (pictures), transcribed, translated and commented texts of letters, as well as photos, posters, audio and video recordings (digital content complemented with metadata).Bibliographic metadata ensure links to digitized first issues of works by both poets: books and articles kept in the Digital Library of the NLL.
The key materials for resource linking by means of annotation are commented texts of letters and work descriptions.Both are structured (by form -both are texts with clearly distinguishable structure) and freely linguistically expressed by content.Only texts in Latvian are annotated, and only one instance of entity was marked in each separate text document.This decision was made keeping in mind one of main objectives of an annotation -the enrichment of plain text data documents with semantically related data by profiting from unambiguous data records held in authority database and other data sources with persistent identifiers.
Another objective of adding an annotation was to identify and mark entities mentioned in texts, such as persons, organizations, events, places, works and concepts, which have at least one of the following data sources to link to: authority or bibliographic record, descriptive data source or full digital representation.A primary source for metadata is the information from MARC21 databases of the NLL: authority database (people, organizations, events, places, concepts), Union Catalogue (manifestations), National bibliography database (manifestations).Links to the Virtual International Authority File (VIAF) for persons, organizations, places, works, and International Standard Name Identifier registry (ISNI) for persons, institutions are also recorded.Wikipedia in multiple languages was used as a descriptive data source.Full digital representations are gathered through links from the NLL Digital Object Management system (DOM) and the Digital Library of NLL (Book Portal and Portal of Periodicals), The European Library and Project Gutenberg.The NLL portal "In search of lost Latvia" provides links to photos of persons and places, including architecture.
The effort ensures that all abovementioned resources have a unique and persistent URIs.For this reason, letters and descriptions of works are hosted in the NLL DOM system, the NLL bibliographic and authority records are imported in RunA project portal.Import of bibliographic and authority records into the project portal is necessary, because they are originally maintained by library information system ALEPH, which allows deleting database records as a normal routine in main catalogue and other databases.On the contrary, NLL DOM system does not allow deleting objects, so the RunA portal can make use of its URIs.
The annotation process highlights various problems even with the most well-known entity class "person".The biggest problem is that the authority data in libraries include mostly persons related to literature.There is a huge "black hole" where scientists, artists, philosophers, socially and politically influential people, and the great part of other significant (or simply mentioned in letters) persons are "lost".If there is a name without an authority record to link to, sometimes other sources, for example, a Wikipedia article, genealogical record or a digital photo could be found, but sometimes it is not possible to provide any link.
Entity class "event" proved itself to be a complex issue.It is not a great problem to link world level events, such as the World War I, which has the subject record in the authority database of the NLL and detailed Wikipedia articles in multiple languages.More local and ordinary events, significant perhaps only to correspondents, as this is in the case of Rainis and Aspazija, are mentioned only in letters themselves.These cases highlight the need for marking simple entities "who", "what", "where" and "when" with the aim to construct a complex entity -"event"."Who" gives an active participant of an event, for example, a name of the theatre, where a play is staged."What" is a name of play (entity class "work"), "where" indicates a place, and "when" is a date.All components together mark an event.
One of the great challenges is creating an annotation of entity "work".Personal correspondence of two writers refers to poems, dramas, novels, articles, monographs, etc., but also staged plays, movies and songs are mentioned.For example, Aspazija in her letter to Rainis on the 23rd of September, 1894, wrote: "Now I will visit Liepaja for a few days.I'm invited to attend "Zaudētas tiesības" staged there by a charitable society.""Zaudētas tiesības" (Lost Rights) is a drama by Aspazija, but there is a staged play with identical title mentioned.Sometimes the publication year and the name of the publisher indicate an RDA Manifestation, which can then be linked to the appropriate bibliographic record.Mostly RDA Works and incidentally Expressions are mentioned.The question is: what must be considered as an appropriate link for those if the data in the NLL are not yet FRBRized?
This issue highlights the necessity to create authority records for RDA Works and Expressions that would be very much appreciated.This question pinpoints the problem of user's expectations when reading annotated texts with links to other resources.Whether the access to authority or bibliographic record should be given to users instead of the digital resource itself, when it is possible?The pilot collection RunA offers a compromise solution -the webpage that was created stores pictures of book covers, bibliographic records and links to the digitized versions of books held in the Digital Library of the NLL.Sometimes only a link to a public digital library is given, when there is no copy of a certain book in the NLL.
Another kind of problems arise when "work" mentioned in the text is, for example, an article, a poem or a short story.None of them have been published separately -they are parts of greater resources such as magazines, newspapers or almanacs.As a result, they do not have bibliographic records to link to.In many cases they have URIs in full-text digital libraries instead, which could be used to point users to the direct page in a periodical or book.
The situation with non-literary works is different and reveals even more difficulties.Movies, lectures, musical compositions and plays, even buildings, monuments and other architectural creations are works with authors and other responsible persons and organizations involved.In most cases these works do not have bibliographic or authority records at all.Occasionally some photo or musical recording could be found.Purely locally known persons involved do not have any authority records in library world and even in Wikipedia.These entities remain unlinked, because there is no URI to point to.

Challenges and limitations of the pilot collection
The pilot collection development team has identified several types of challenges for organisation and identification of bibliographic entities during the pilot project:  The scope of the NLL authority database is not sufficient for heritage projects of crossorganizations as it does not cover all kind of essential entities;  As the NLL catalogue and databases are not yet FRBRized, there is a lack of Work/Expression authority records;  Drawbacks in identification and linking of Time-span/Event entities and Place entities with geographic location/co-ordinates.For example, these entities are not used to create RunA timeline;  Limitations in identification and linking of abstract entities/subjects;  The same types of key entities (works, expressions, manifestations; agents: persons, institutions; concepts, places, events) are identified in the object metadata of the collection and also in annotated textual documents.Therefore the necessity to use novel methods to identify and reflect simple or hierarchical relationships between different entities is realized.
The limitations of the pilot project caused by technical solutions are reflected in the article by U. Bojars (Bojars 2016, 21-26).
The main problem related to the identification of entities was that information on entities has been organised in MS Excel tables (instead of a separate database) classifying:  entity type (class);  preferred form of entity (from the NLL authority and bibliographic data as well as free chosen forms in Latvian);  identification of internal (the NLL database or DOM system) or external (VIAF, Wikipedia, etc.) sources.
Identification of external entities/resources is still problematic.In the current development stage mostly local URIs are provided to identify entities.

RunA main groups of entities and the RDA rules
RDA contains all the entities that are of key importance in RunA collection.These named entities are serving as access points in Semantic Web applications, identifying and connecting related entities."To identify and to relate entities" are two functions realized by RDA (Bianchini and Guerrini 2016, 88).
Some research has been done by the RunA team to compare data models FRBR and IFLA LRM with RDA and BIBFRAME.The issues related to RDA Work, Expression and Manifestation level entity identification and linking are of key importance, for instance, to distinguish between different usage of entity "Faust" (by Goethe) in the correspondence of Rainis and Aspazija.When exactly do both mean "the intellectual or artistic creation", "the physical embodiment of an expression of a work" or just "Faust" as the literary hero?For example, Aspazija in her letter to Rainis in the middle of June, 1897, wrote: "I will edit Faust as much as I can, and will send it to you tomorrow.""Faust" mentioned here is a German to Latvian translation of Goethe's "Faust" by Rainis and Aspazija.Therefore, this is an FRBR expression.
RunA project uses RDA defined entities -Works (works written by Rainis un Aspazija, their translated works, works that are inspirational sources; related works, which include their works; works by other authors that are based on literary works by Rainis and Aspazija), Persons (persons related to Rainis and Aspazija; real, historic or fictional persons mentioned in their works; literary and mythological characters, like Undine), Institutions (publishing houses, commercial organizations, cultural and educational institutions), Concepts (specific example: bylina -Russian epic poem; abstract example: nirvana), Events (local example: Latvia, German occupation (1941)(1942)(1943)(1944); global example: the First World War), Places (geographical -the Baltic sea; historical -USSR, Kievan Rus').
Named entities in RunA annotations are connected to bibliographic data created in the NLL, and can be identified using authority data.Identifiers are local authoritative data set numbers, VIAF identifiers and ISNI numbers for people and organizations.Authority data are created for Persons, Families, Institutions, Places, Events, Concepts.As was concluded from the pilot project, a huge setback for the project is the fact that authority data for Works and Expressions have still not been created.It means that it is currently not possible to identify Works, their Expressions, or to show their interrelations and related Works and Expressions.Difficulties with person identification, in turn, are often based on the fact that personal names tend to correlate to personal names, works or place names, which are not identified.Collection, for instance, includes several annotations for the name "Aspazija", which is identified as different persons (the poetess Aspazija and the Ancient Greek feminist Aspazija), as well as with some Work titles (a play "Aspazija" by a poetess and a playwright Aspazija, which is based on the novel "Aspasia" by Robert Hamerling).There is also a staged play "Aspazija", based on the play "Aspazija" written by Aspazija, which can be considered as another Expression of this Work.If RDA record for the play "Aspazija" is made, it would have to be connected to the Work record for the novel "Aspasia" written by Robert Hamerling, by explaining their relations as interconnected works with an attribute "inspired by "Aspasia" -a novel by Robert Hamerling" and "Inspiration for "Aspazija" -a play by Aspazija".It has to be concluded, that even though it is not possible to show reciprocal relationships between related Works in MARC bibliographic records, it can still be done by writing the name of the author and the title of his  Unfortunately, records created in RIMMF cannot be exported to Aleph system.Names of Geographical locations can have multiple meanings and connect to Works, Persons, Institutions and Events.For instance, project annotations connect river Daugava to the poem "Daugava" by Rainis, which in turn is connected to a specific historical Time span -National awakening, liberation wars and early formation of Latvian state (1915)(1916)(1917)(1918)(1919).A play with the title "Daugava" based on this poem was staged later (derivative work).M. Brauns, inspired by the poem, wrote a song "Sun, thunder, Daugava" (inspired work), which links directly to the event "Song festival".Daugava (location -a river) is also linked to the abstract notion, which only Latvian public can comprehend -the "River of fate".
Some mutually related entities are still left unidentified and the relationship hierarchy between these elements is yet to be discovered.A play "Fire and night" by Rainis can be called as a bright example of that.Rainis used an epic poem by Andrejs Pumpurs "Lāčplēsis" ("Bearslayer") as the base for his work "Fire and night".Folk tales and legends about Lāčplēsis in turn inspired the work "Lāčplēsis" by Pumpurs.This exact work of Rainis is thus a derived and inspired work.Relationships between Works can be demonstrated using RDA entity hierarchy model.

NLL authority data and RunA entities
Currently movement from authority control in the direction of identity management can be traced in the sphere of library data development.This change is accompanied by the philosophy change and change in the scope of authority data."Large-scale systems will require a shared store for linked data triples with facilities for Uniform Resource Identifier (URI) management to evolve authority control into entity control" (Dunsire, Fritz and Fritz 2016, 4).
In the case of RunA, information on entities has initially been organised in MS Excel tables.This is just a short-term solution, while an implementation of a special data store for information on entities, with a feasibility to re-use it in other applications is in progress.
"Since 2000 the NLL has maintained the authority records database.It serves as a national bibliographic resource for personal names, corporate bodies, geographic names, and subjects.In summer 2013 authority records for personal names and corporate bodies created by the NLL were successfully incorporated in the Virtual International Authority File (VIAF)" (Goldberga et. al. 2014, 215).
VIAF publishes authority data of the NLL as Linked Data, secures it being linked to other countries' authority data, Wikipedia, and other resources.At the beginning of 2017, the NLL authority database had approximately 185,000 authority records for persons and institutions with VIAF ID attached, and about 135,000 of them with ISNI.
Starting from 2016, the NLL has been creating authority data for persons, institutions, families, and geographic names, which fully match current RDA rules; therefore authority records contain much more detailed information related to entities, such as places, activity fields, occupation/profession, gender, language, historical data and other attributes to identify an entity.In accordance with RDA standard authority data allow to reveal relations between persons, families, and institutions, as well as person, institution, and family relations with other FRBR entities -Work, Manifestation or Expression, that match the requirements of Linked Data and Semantic Web.
No authority data for Works have yet been created at the NLL, therefore the experience of other libraries and projects on the extraction of information on Work entity from bibliographic data was being researched in order to understand the automated FRBRization possibilities and entity matching principles.Here are some examples: National Library of Spain -datos.bne.es;National Library of France -data.bnf.fr;Syrtis (Switzerland), etc.
At the beginning of 2017, mapping was prepared to extract Work authority data from the NLL's bibliographic data (MARC21).
The future plans of the NLL include the creation of authority data collection of national importance that would use the principles of Linked Data.Although at present, navigation from our authority records to VIAF data clusters (by using VIAF ID connected to authority data) is the only way to make our data a part of Semantic Web.
Authority data in RunA collection were chosen to be the basis for identification of named entities during the annotation process, as well as the crossing points for interlinked information based on semantics.
For example, one of the annotations in RunA collection is dedicated to the collection of poems "Tie, kas neaizmirst" (Those who don 't forget, 1911), and it provides the information about some of the mottos of foreign authors included in the edition.One of the authors, whose motto is used in the collection, is Mazzini, Giuseppe (1805-1872) -Italian politician, journalist, activist for the unification of Italy.In this annotation he was identified using the NLL person authority record.While the authority data of the NLL help to identify this person to those who are not familiar with Latvian language, as it is written in the original transcription with the addition of years of life, only the VIAF ID secures the connection to other related information.The annotation to the poem of J. Rainis "Ave sol" (Hail the Sun, 1910) the name of which includes an ancient Roman greeting, is also included in the collection.This annotation also provides the information about one of Latvian cultural symbols, the choir "Ave sol" (Corporate Body), which carries the name of the poem.This entity was identified with the NLL authority record, and the data included inside help to identify and provide more exhaustive information about the choir, but the VIAF ID secures the link to other related resources.
Authority data for geographical place names created at the NLL have not yet been included to VIAF.
The Place entity (building address) is often identified as a specific Corporate Body (institution), even though different institutions can occupy that building at different time.It would be better to create authority data by treating buildings like objects, while additionally providing information about their architects, builders, owners, related institutions, their history, and social significance.This information is an essential part of national cultural and historic heritage, which is absent from data at present.
The entity mentioned in several annotations of the collection -the abstract object "Glass Mountain" can be looked at in the context of the play by Rainis "Zelta zirgs" (The Golden Horse), as well as the building of the National Library of Latvia designed by Gunnar Birkerts.The architect Birkerts was inspired by the Glass Mountain depicted in a play of Rainis, as the symbol of National awakening.The object Glass Mountain and the abstraction of attaining spirituality by "climbing the Glass Mountain" are connected to both the "National Library of Latvia" as an institution, and with the design of the building by Birkerts.Birkert's project of National Library of Latvia itself can be considered as Work, and according to RDA needs to have a Work record, which would include an author -Birkerts, Gunārs (Agent); title and its variations (National Library of Latvia, Glass Mountain, Castle of Light, Lighthouse, etc.); date and place of creation of this Work (project); connection to other Works: "Zelta zirgs" (The Golden Horse) by Rainis, poem by Auseklis "Gaismas pils" (Castle of Light), song by Jazeps Vitols "Gaismas pils" (Castle of Light).
Talking about the entity "Lāčplēsis" (Bearslayerliterary hero, fairy tale character), it is important not only to identify its relations to different Agents but also to show the mutual relations between them.It is possible if we are using the environment, which allows to structure data in RDA entity levels (FRBR WEMI).RIMFF environment allows creating data structure that matches the structure of Linked Data.It is possible to create separate Work, Expression, Manifestation, and Item entries as well as authority records by linking them together to show mutual relations between these entities.
Identification of entities in unstructured data, as in letters and annotations in RunA, is further complicated by the natural polysemy of language inherent in this data, acronyms, the meaning of which is clear only to authors of the letters, the usage of nicknames and signs, as well as inserting the information in other languages.In addition to that, vague expressions related to time or place like 'three days ago', 'this year', 'in the sea side', 'the Craftsmen's hall', etc., and the necessity to involve well educated and knowledgeable specialists on Latvian culture and literature in the process of identification of entities, makes the process even more cumbersome.
Person identification process can also be tricky because of the usage of natural sequence 'name, surname' in the letters, instead of 'surname, name' sequence of authority records.Correspondence also includes only names, only surnames, other words like 'uncle'.Therefore involvement of qualified specialist is needed here as well.
Latvian language is one of inflected languages, where grammatical categories of gender, number or conjugation in conjugable parts of speech are expressed with conjugable word endings.Word ending depends on the context, thus is objected to change.The usage of conjugated words in unstructured data also complicates the process of entity identification, as in structured data like authority or bibliographic data words are usually put in Nominative singular or plural.
In the process of identifying concept entities in unstructured data, one should really keep in mind the natural ambiguity of the language and that by using subject authority records for concept identification, to denote a concept in unstructured data, different synonym can be used.For instance, the concept of 'feminism', its synonym form can be 'emancipation of woman'.
In order to serve better for purposes of entity identification for library created authority data, authority forms would have to be supplemented with a larger amount of concepts, place names and synonym forms.Additionally, perhaps, person authority records should include 'name surname' form used in natural language.Furthermore, authority entries should be created for Latvian people and institutions not connected directly to literary process, although significant in other contexts.

Potential of RDA and entity identification with URI
Research done by the RunA collection development team reflects some strengths and weaknesses of RDA.
RDA strengths:  Allows to build an entity relationship network including reciprocal relationships between entities identified with global URI (Open Metadata Registry);  There are defined URIs for relationship designators in RDA Registry (language independent), e.g., has founder -P50029;  Compatible with FRBR and IFLA Library Reference Model (LRM); developing under the influence of LRM; supports hierarchy of entities;  Reduces data duplication;  Is a candidate to become a data content standard for memory institutions;  Allows data in different formats (MARC, DC, etc.);  Supports the use of other de facto controlled vocabularies, ontologies.

RDA weaknesses:
 RDA arrives from the existing library standards and still lacks the depth that the crossdomains possess;  Problems with relationship designator labels in Latvian (impact of conjugations); Relationship designator label "Contained in (work)" is a phrase designed for user interface.Exact translation in Latvian -"Ir daļa no (darbs)" does not correspond to the correct use of conjugations in Latvian (correct form: "Ir daļa no darba").From the user perspective only first part of label "Contained in (work)" -"Contained in" is essential.Other part "(work)" is important for structuring and importing data from MARC to other data environments, so it could be coded, for example, by using CURIE (Compact URI) from RDA Registry.
Example:  Problem in re-using other data.Too much subjectivity in the depth/level of detail; For example: RDA does not specify whether CD provides information on every Work it contains or the whole CD in general (depends on the local decision).
 The necessity to define new relationships that could appear/emerge; It is possible that the necessity to define new relationships can arise.For example: relationship between the entity 'building' and entities 'architect', 'constructor', 'owner'.
 To reflect entities mentioned in content of all documents RunA team would appreciate additional relationship in RDA "is mentioned" (for instance, another author, work, concept, etc. is mentioned in content of Work).
Comparing to use of subjects in similar cases relationship "is mentioned" would better serve to the aims of RunA because it is not possible always to find terms from controlled vocabularies.Linked Data representations of RDA components in RDA Registry "allow developers to improve the interoperability of data produced by RDA with a wider base of international standards."(Dunsire 2016, 17).
It is possible to declare a unique RDA URI while creating MARC21 authority data for persons, families, institutions, places, items, and events.These are possible to obtain from vocabulary embedded in Open Metadata Registry and add to authority records in 5XX reference fields by using |4 Relationship code subfield.

Identification of abstract entities of RunA
To move away from isolated data there is a significant need to deal persistently with identification of abstract entities like "feminism", "women's rights".For example, linking Aspazija's dramatic works entities with related entities in the correspondence to Rainis.
Applying identification, semantic linking and networking of entities in case of unstructured data sets, there is sometimes a necessity to identify and semantically link concept and event entities.Controlled vocabularies for subjects or ontologies would be the best solution to fulfil the "task of navigating in the world of recorded knowledge" (Bianchini and Guerrini 2016, 2).
NLL uses its own created Subject Headings (NLL-SH) in Latvian language for thematic access, which is a list based on the adapted Library of Congress Subject Headings (LCSH) subject indexing language.
"Most of NLL-SH records were adapted from LCSH by translating preferred labels into Latvian while other records, specific to the NLL and Latvia, were introduced without having a matching LCSH concept.The NLL fully complies with the LCSH system structural regularities -hierarchical relationships, associative relationships and syntax.Library of Congress original subject headings are included as See from references, thus standardizing the translation and providing for international data retrieval" (Stūrmane, Eglīte and Jankevica-Balode 2014, 20-21).
NLL-SH authority data in RunA collection were used to identify different abstract entities, concepts and events, like "feminism, alchemy, dragons, Christmas, World War I, 1914-1918", in annotations and comments.
Subject authority records secure entity identification of these concepts and events, as they also contain synonym forms of these concepts and events in Latvian language, chronological data for events, as well as concept names in English, which became possible due to Library of Congress original subject headings.While NLL-SH transition to Linked Data has not been completed, further semantic linking is not possible.Different solutions for NLL-SH linking with subject controlled vocabularies in other languages, especially those represented in form of Linked Data, are being considered and researched.
One possible solution for that is through a connection with LCSH by using Library of Congress original subject headings attached to NLL-SH authority data.U. Bojārs, the researcher of the NLL in his paper "Linking Library Data for Quality Improvement and Data Enrichment" states: "In the case of taxonomies, such as library authority data, linked records from both taxonomies may contain labels in different languages and these records can be enriched by copying labels across datasets, facilitating creation of multilingual taxonomies.Authority data records may contain links, both internal and external, that can be a valuable resource for data enrichment.Once a link between NLL-SH and LCSH records is established, NLL-SH records can be enriched with links to authority records from the National Library of France and the German National Library that are included in the LCSH dataset".(Bojārs, Žogla and Eglīte 2015, 187).
The NLL is currently in its transitional period from NLL-SH to a subject indexing system similar to FAST (Faceted Application of Subject Terminology), the syntax of which would be simplified and more suitable for the environment of World Wide Web.At this stage form/genre facets are being created.After this transition is over, it will be possible to make a bond both with FAST and corresponding and interconnected LCSH concepts by using English concepts of FAST, which will be added to the NLL subject authority data.Therefore, the system would be able to provide concept identification and multilingualism.
Transition to a system similar to FAST would solve the problem with entity Time-span identification and semantic linking, because the NLL currently does not create chronological term authority data and they are not being used as separate access points in bibliographic records as well.Chronological dates as subdivisions appear only in authority records for persons, institutions, topical subjects or events.In future it will be necessary to create Chronological Term authority records corresponding to FAST, which would serve as identifiers for Time-span entity and secure its semantic links.
Bearing in mind that concept level identification needs in Linked Data collections will only increase, the NLL experts are considering possible collaborations with developers of Finnish KOKO ontology http://seco.cs.aalto.fi/ontologies/koko/.The potential bases for cooperation -English concepts of LCSH used in the NLL authority records.
Aspazija was concerned by her contemporaries as a prominent defender of feminist ideas in Latvia, which found reflection in annotations and commentaries of the collection.To identify the entity for the concept of "feminism" RunA contains NLL-SH authority record, but by linking English concepts of LCSH of this concept with multilingual Finnish KOKO ontology, identification with URI from this ontology, accessibility in multiple languages as well as access to other related resources becomes possible.defines properties and subproperties (expressing relations) which are identified with URI.This is perfectly suitable for linking entities in collection such as RunA.The only drawback is the "human" part of representing data -RDA properties must be translated in RunA interface languages.
One of RunA collection main tasks is an evaluation of new library and web standards and re-use of bibliographic and authority data in Web environment linking with entities from other metadata sets and objects.Why?Because currently in majority of cases the NLL data are the final destination for our users: they do not allow contextualisation and navigation to full texts, objects or other data sets, i.e. it is not possible to navigate from "strings to things".
RDA provides memory institutions with novel rules to identify detailed relationships between entities: simple and hierarchical.However it will take a long time to adapt and implement these rules in practice.
Future objective of this collection is to reflect objects, their attributes and relationships using key entity classes, subclasses and the hierarchy of their relationships according to IFLA LRM semantic levels.
RunA Linked Data interface and open datasets are published and perhaps some researchers may find them to be useful.
Ongoing work on development of special annotation tool and separate data store for entities including all related contextual information and unique identifiers is still in-progress.

Figure 1 .
Figure 1.RDA key entities and their relationships with FRBR, IFLA LRM and BIBFRAME.

"
From the April 2014 on, the NLL uses new RDA MARC21 fields 370 to 378 to create new and enrich existing authority records for persons, corporate bodies and family names.Along with the implementation of the new RDA MARC21 fields and participation in the VIAF, MARC21 field 024 (Other Standard Identifier) and MARC21 field 035 (System Control Number) are added to the NLL authority records, providing International Standard Name Identifier (ISNI) and the VIAF ID" (Ibid, 215-216).

Figure 8 .
Figure 8. KOKO page for concept "feminism". 4 This structure still does not create Linked Data, and is only capable of demonstrating relations between the two Works visually within a MARC record.In a similar way, MARC environment is not suitable for demonstrating relationships between Work and its multiple Expressions and Manifestations within the structure of Linked Data.A good example of that being the collection of poems by Rainis -It is also difficult to show relationships of this Work with related entities like inspired works (oratorio, symphony, song cycle) and CorporateBody (chamber choir "Ave Sol") in the MARC structure.Work record that includes the following information about contained works in general would be necessary: author Rainis, Jānis, 1864-1929, title "Dagda's sketch books" (alternative title -"The novel of Dagda"), work completion date, place or places where the literary work was written.Furthermore Work records for every poem is necessary as well, since all of them can be considered as separate Works (Aggregate).Currently, MARC and ALEPH system allow creating one record with Work, Expression, Manifestation and Item attributes included, and not a separate bibliographic record for each FRBR WEMI level.It is not possible to demonstrate neither hierarchical levels within the RDA model, nor related entity relations as a Linked Data model.Although by using RIMMF -an RDA training tool, it is possible to create an individual entry for each WEMI level and link them together to show relationships between those levels and entities included.
Work in subfield |i together with an appropriate indicator within field 700 in MARC21.1000#|a Aspazija, |d 1865-1943 |4 aut 245 10 |a Aspazija : |b Sen-Helladas drāma / |c Aspazija.7001#|i Inspired by: |a Hamerling, Robert, |d 1830-1889.|tAspasia."AveSol"(Work).It is translated into a number of other languages which count as different Expressions of this Work and only a preferred title and an authority record for author connect their bibliographic records, with Work level absent.1001#|a Rainis, Jānis, |d 1865-1929 |4 aut 240 10 |a Ave sol! |l Lietuviešu valodā.1001#|a Rainis, Jānis, |d 1865-1929 |4 aut 240 10 |a Ave sol! |l Krievu valodā.Relationship hierarchy of entities in RunA and RDAProblems tend to arise upon trying to identify the Place entity, when the name of the place coincides with other entities (e.g.personal name).The city of Dagda (Dagda, Dagda municipality) shares the name with one of characters in the works of Rainis.If Person Dagda does not have its own authority record, it is not a simple task to identify him.Dagda as a fictional character is connected to the series of works by Rainis -"Dagda's sketch books", which includes a number of poems, each having an individual title.Without Work authority record it is difficult to identify separate parts of this Aggregate Work.It is possible, though, to show these relations inside the bibliographic records in MARC environment with the help of field 700, where the name of the author, the title and the relationship designator label "(Part of (Work))" is specified under the subfield |i.

Table 1 .
Entities essential to RunA collection and which are the backbone of authority data have already been defined in FRBR: Work, Expression, Manifistation, Person, Institution, Concept, Event, and Place.FRAD defines one more entity -Family.Upon evaluating global models, standards and in OMR published namespaces that are of interest to the NLL, the project team came to a conclusion that the only data standard, the elements/entities of which are identified with globally recognized URI and used by the NLL (FRBR is not used) is RDA: Key entities in RunA comparing to FRBR and RDA corresponding entities.