Wikidata:Property proposal/interpretation
interpretation edit
Originally proposed at Wikidata:Property proposal/Generic
Description | interpretation of sentence sense in natural text in rdf format |
---|---|
Data type | Item |
Domain | property |
Example 1 | Genesis 1 Line 1 (Q91393642) → Genesis 1 Line 1 Interpretation (Q91460154) |
Example 2 | Genesis 1 Line 2 (Q91393643) → Q91554315 |
Example 3 | Genesis 1 Line 3 (Q91554569) → Q91554811 |
Motivation edit
I want to create rdf interpretations for popular books (like Bible) and than use this interpretations as dataset for machine learning NLP tasks.
I also think about properties like statement, subject (Q164573), predicate (Q179080), object (Q175026), all types of adverbial (like Time adverbial (Q25592429)) to use in interpretation
Deepsaged (talk) 05:42, 22 April 2020 (UTC)
Discussion edit
- Comment @Deepsaged: Your examples don't provide any information - they are all just new items in Wikidata. Can you give us some actual details of what you're trying to do here, at least some sort of background where somebody has done something like this before? I don't personally understand why the values for this property need to be Wikidata items, nor am I particularly enthusiastic about creating new items for every line of text in even 1 book. ArthurPSmith (talk) 17:42, 23 April 2020 (UTC)
- Comment @ArthurPSmith: Hello, thank you for the answer.
Main idea is to create relations from tokens (lexemes) in text to wikidata items
I hope if it will be done for big amount of text it will be possible to train some machine learning models to do it automatically
So we can say that computers can in some way "understand" or create "interpretation" on natural text. In this case under “interpretation” a mean to create correct link from text tokens to wikidata entities and for example solve homonyms
I have the sketch of the first line interpretation in RDF. It is not well formatted N-Triples, just main concept
Please look:
@prefix bible: <example-prefix-for-bible-interpretation> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix ds: <example-prefix-for-not-existing-entities-and-properties> . # ds is DeepSage :-)
bible:book/gen/chapter/1/line/1 dcs:interpretations bible:book/gen/chapter/1/line/1/interpretations . #here we point that one text line can have many interpretations
bible:book/gen/chapter/1/line/1/interpretations rdf:type rdf:Bag .
bible:book/gen/chapter/1/line/1/interpretations rdf:_1 bible:book/gen/chapter/1/line/1/interpretation/1 . #this is first, and the only one interpretation for current text
bible:book/gen/chapter/1/line/1/interpretation/1 rdf:type rdf:Seq . # here we say that interpretation is array of statements
bible:book/gen/chapter/1/line/1/interpretation/1 rdf:_1 bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 . # and here we have our main statement
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 rdf:type rdf:Statement .
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 rdf:subject God in Christianity (Q825) . # sentence subject
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 rdf:predicate Q91463220 . # sentence predicate
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 rdf:object heaven (Q4489450) .
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 rdf:object Earth (Q2) .
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 temporal adverbial (Q25592429) bible:book/gen/chapter/1/line/1/interpretation/1/adverbial/1 .
bible:book/gen/chapter/1/line/1/interpretation/1/statement/1 temporal adverbial (Q25592429) bible:book/gen/chapter/1/line/1/interpretation/1/adverbial/2 .
bible:book/gen/chapter/1/line/1/interpretation/1/adverbial/1 <ds:context-relation> Q90219924 .
bible:book/gen/chapter/1/line/1/interpretation/1/adverbial/1 <ds:context-space-part> beginning (Q529711) .
bible:book/gen/chapter/1/line/1/interpretation/1/adverbial/2 grammatical tense (Q177691) past tense (Q1994301) .
Deepsaged (talk) 18:56, 23 April 2020 (UTC)
- @Deepsaged: I like the idea. But you don't need a property to link a work to its corresponding interpretation, because each interpretation can link to the interpretated text with main subject (P921). You can get all interpretations of a given item using SPARQL. Check what I've tried to do in Genesis 1 Line 1 Interpretation (Q91460154). However, in order to model the example you wrote above, we will need new properties. --Tinker Bell ★ ♥ 03:05, 24 April 2020 (UTC)
Notified participants of WikiProject Linguistics
- @Tinker Bell: Thank you for the example! Before creating property proposal I searched properties by names like "sense" "interpretation" "meaning" and was not able to find one. Now I see you example and it is cool that we can use existing properties for my task. Now I will think about other elements of model from my example. Deepsaged (talk) 05:31, 24 April 2020 (UTC)
- @Tinker Bell: I moved your example to new item Genesis 1 Line 1 Interpretation Statement 1 (Q91718922) because I think we need general structure for cases with more than one statement in sentence. Please look. Deepsaged (talk) 06:04, 24 April 2020 (UTC)
- @Deepsaged: I think, it is an interesting idea to describe the bible this way. However I do not think Wikidata is the right database for this project nor do think Wikibase is the right tool for it. --CamelCaseNick (talk) 20:58, 25 April 2020 (UTC)
- @CamelCaseNick: But why not wikidata? May be I do not know something or there is better platform for this task? Deepsaged (talk) 06:50, 26 April 2020 (UTC)
- Not done no support, it's unclear if WD is suitable for this. --- Jura 09:43, 15 July 2020 (UTC)