Conllu Parsing and Searching

From Apertium
Jump to navigation Jump to search

Parse and Search through a conllu file

Searching is as follows:


Tree search: The '{' character

If you would like to search with a tree(i.e you want to search for a word with a HEAD value or word), you create your search with a '{'. Then, between the words you are searching for a relation between add a '>'. You can also use and '<' if you are searching for a word that is a dependent of another word. This, the '<', will find the dependent word. For instance, if you wanted to see when 'have' did action to 'clue' (i.e. I have no clue') you could do it like this:

For example python conlluparse.py "text.conllu" '{have>clue' might output:

Token: 2, Form: have, Lemma: have, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = 2, Sentence:  I have no clue .

If you wanted, you could also be more specific or ambigious with your searches. When you specify these arguments, you also need to make sure that you concatenate "Form=" with the word you are searching for. When you have nothing specified on one side, you need to add 'none=none' to that side. For instance if you wanted to find if something was a dependent of 'have', you could do:

{none=none<form=have

When searching with attributes (i.e UPOSTAG), you could do this like:

python conlluparse.py "text.conllu" '{upostag=verb, form=have>form=clue' which may output:

Token: 2, Form: have, Lemma: have, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = 2, Sentence:  I have no clue .

You can search with any of these tags - upostag, xpostag, lemma, or deprel. You would do this by just putting the tag name + and '=' and then the actual value. Concatenate the tag an '=' and the value like upostag=noun' or 'lemma=clue' or @.

You can also specify attributes instead of 'form=clue' such as 'upostag=noun'

Token: 2, Form: have, Lemma: have, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = 2, Sentence:  I have no clue .

Now, instead, if you search with python conlluparse.py "text.conllu" '{form=clue<none=none' , it will print:

Token: 4, Form: clue, Lemma: clue, UPOSTAG: NOUN, HEAD: 2, DEPREL: obj, # sent_id = 2, Sentence:  I have no clue .

The tree searches essentially combines all the searches terms. You can search for a word with a specific deprel or upostag like

{@root, upostag=noun>none=none

You can search for relationships like the ; character:

{@nsubj>upostag=noun

You can search for a plain word like:

{form=have>none=none

This will output the same thing as what the other searches would do. Use other searches to distinguish between what you are searching for.

Find Features: The ':' character

If you would like to search for a deprel or upostag and a feature in a word, you would start your search with a ':' and encapsulate your search with '[]'

For instance if you wanted to search for a copula and past feature you would do

python conlluparse.py "text.conllu" ':[cop, past]' which may output:

'Token: 3, Form: болғаныма, Lemma: бол, UPOSTAG: AUX, HEAD: 2, DEPREL: cop, # sent_id = akorda-random.tagged.txt:158:2829, Sentence:  Мен осында болғаныма қуаныштымын қуанышты мын .'

Relationships: The ';' character

If you would like to search with a relationship(i.e nsubj relation to another node that has a noun POS), you would start your search with a ';'. You would then type a deprel tag followed by a colon and then a part of speech. The second term(the one after the ';') can be these tags : lemma or the word id_name. You would use to search for a word with nsubj relationship with a noun:

python conlluparse.py "text.conllu" ';nsubj:noun' Would output:

'Token: 8, Form: жүзімдік, Lemma: жүзімдік, UPOSTAG: NOUN, HEAD: 6, DEPREL: conj, # sent_id = Шымкент.tagged.txt:8:216, Sentence:  Тау етегінде өзен бойындағы алқаптарда егіншілік пен жүзімдік  ал көгалды таулы жайылымдарда - мал шаруашылығы дамыған .'

Examples

python conlluparse.py "text.conllu" ':[aux, pres]' This is how you would run the program with the ':' Could output:

Token: 5, Form: жатырсыздар, Lemma: жат, UPOSTAG: AUX, HEAD: 4, DEPREL: aux, # sent_id = akorda-random.tagged.txt:44:775, Sentence:  - Сіздер осында тұрып жатырсыздар  ал Астанада жұмыс істейсіздер .

python conlluparse.py "text.conllu" ';nsubj:verb' This is how you would run the program with the ';' Could output:

Token: 7, Form: секіреді, Lemma: секір, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = Ер_Төстік.tagged.txt:22:380, Sentence:  Қуып келе жатқан Төстік Төстік те өрмектен секіреді .

python conlluparse.py "text.conllu" "{none=none>none=none, @obj" This is how you would run the program with the ';' Could output:

Token: 6, Form: іліп, Lemma: іл, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = Ер_Төстік.tagged.txt:23:396, Sentence:  Сөйткенде Төстіктің бір бақайы өрмекті іліп кетеді .

python conlluparse.py "text.conllu" '<іліп' This is how you would run the program with the '<' Could output:

Token: 6, Form: іліп, Lemma: іл, UPOSTAG: VERB, HEAD: 0, DEPREL: root, # sent_id = Ер_Төстік.tagged.txt:23:396, Sentence:  Сөйткенде Төстіктің бір бақайы өрмекті іліп кетеді .