Spaces:

AUEB-NLP
/

The-Greek-NLP-API

Sleeping

File size: 14,407 Bytes

from typing import Dict, List

from fastapi import FastAPI, HTTPException, Query
from fastapi.responses import RedirectResponse
from gr_nlp_toolkit import Pipeline
from pydantic import BaseModel, Field

app = FastAPI(
    title="The Grεεk NLP API 🇬🇷",
    description="State-of-the-art API for Greek NLP tasks including Greeklish to Greek conversion (G2G), Named Entity Recognition (NER), Part-of-Speech (POS) tagging, and Dependency Parsing (DP). The API is powered by the Grεεk NLP Toolkit ([https://github.com/nlpaueb/gr-nlp-toolkit/](https://github.com/nlpaueb/gr-nlp-toolkit/)), which is also available via PyPI (`pip install gr-nlp-toolkit`). ",
    version="1.0.0",
    contact={
        "name": "Natural Language Processing Group - Athens University of Economics and Business (AUEB)",
        "url": "http://nlp.cs.aueb.gr/",
        "api_author": "Lefteris Loukas",
    },
)

# Instantiate the Pipeline
nlp_pos_ner_dp_with_g2g = Pipeline("pos,ner,dp,g2g")


# Pydantic models for responses
class G2GOutput(BaseModel):
    greek_text: str = Field(
        ...,
        example="η θεσσαλονικη ειναι ωραια πολη",
        description="Converted Greek text",
    )


class NERItem(BaseModel):
    token: str = Field(..., example="αργεντινη")
    ner_value: str = Field(..., example="S-ORG")


class POSItem(BaseModel):
    token: str = Field(..., example="μου")
    upos: str = Field(..., example="PRON")
    morphological_features: Dict[str, str] = Field(
        ...,
        example={
            "Case": "Gen",
            "Gender": "Masc",
            "Number": "Sing",
            "Person": "1",
            "Poss": "_",
            "PronType": "Prs",
        },
    )


class POSResponse(BaseModel):
    pos_results: List[POSItem] = Field(
        ...,
        description="Part-of-Speech tagging information",
        example=[
            {
                "token": "μου",
                "upos": "PRON",
                "morphological_features": {
                    "Case": "Gen",
                    "Gender": "Masc",
                    "Number": "Sing",
                    "Person": "1",
                    "Poss": "_",
                    "PronType": "Prs",
                },
            },
            {
                "token": "αρεσει",
                "upos": "VERB",
                "morphological_features": {
                    "Aspect": "Imp",
                    "Case": "_",
                    "Gender": "_",
                    "Mood": "Ind",
                    "Number": "Sing",
                    "Person": "3",
                    "Tense": "Pres",
                    "VerbForm": "Fin",
                    "Voice": "Act",
                },
            },
            {
                "token": "να",
                "upos": "AUX",
                "morphological_features": {
                    "Aspect": "_",
                    "Mood": "_",
                    "Number": "_",
                    "Person": "_",
                    "Tense": "_",
                    "VerbForm": "_",
                    "Voice": "_",
                },
            },
            {
                "token": "διαβαζω",
                "upos": "VERB",
                "morphological_features": {
                    "Aspect": "Imp",
                    "Case": "_",
                    "Gender": "_",
                    "Mood": "Ind",
                    "Number": "Sing",
                    "Person": "1",
                    "Tense": "Pres",
                    "VerbForm": "Fin",
                    "Voice": "Act",
                },
            },
            {
                "token": "τα",
                "upos": "DET",
                "morphological_features": {
                    "Case": "Acc",
                    "Definite": "Def",
                    "Gender": "Neut",
                    "Number": "Plur",
                    "PronType": "Art",
                },
            },
            {
                "token": "post",
                "upos": "X",
                "morphological_features": {"Foreign": "Yes"},
            },
            {
                "token": "του",
                "upos": "DET",
                "morphological_features": {
                    "Case": "Gen",
                    "Definite": "Def",
                    "Gender": "Masc",
                    "Number": "Sing",
                    "PronType": "Art",
                },
            },
            {
                "token": "andrew",
                "upos": "X",
                "morphological_features": {"Foreign": "Yes"},
            },
            {
                "token": "ng",
                "upos": "X",
                "morphological_features": {"Foreign": "Yes"},
            },
            {"token": "στο", "upos": "_", "morphological_features": {}},
            {
                "token": "twitter",
                "upos": "X",
                "morphological_features": {"Foreign": "Yes"},
            },
            {"token": ".", "upos": "PUNCT", "morphological_features": {}},
        ],
    )


class DPItem(BaseModel):
    token: str = Field(..., example="προτιμω")
    head: int = Field(..., example=0)
    deprel: str = Field(..., example="root")


class DPResponse(BaseModel):
    dp_results: List[DPItem] = Field(
        ...,
        description="Dependency Parsing information",
        example=[
            {"token": "προτιμω", "head": 0, "deprel": "root"},
            {"token": "την", "head": 4, "deprel": "det"},
            {"token": "πρωινη", "head": 4, "deprel": "amod"},
            {"token": "πτηση", "head": 1, "deprel": "obj"},
            {"token": "απο", "head": 7, "deprel": "case"},
            {"token": "την", "head": 7, "deprel": "det"},
            {"token": "αθηνα", "head": 4, "deprel": "nmod"},
            {"token": "στη", "head": 9, "deprel": "case"},
            {"token": "θεσσαλονικη", "head": 4, "deprel": "nmod"},
            {"token": ".", "head": 1, "deprel": "punct"},
        ],
    )


# API endpoints
@app.post("/g2g", response_model=G2GOutput, summary="Convert Greeklish to Greek (G2G)")
async def greeklish_to_greek(
    text: str = Query(
        ...,
        description="The Greeklish text to convert",
        example="H thessaloniki einai wraia polh",
    ),
):
    """
    The G2G (Greeklish-to-Greek) endpoint takes Greeklish text (Greek written with Latin characters) as input and transliterates it to Greek text.
    """
    try:
        greek_text = " ".join(
            [token.text for token in nlp_pos_ner_dp_with_g2g(text).tokens]
        )
        return G2GOutput(greek_text=greek_text)
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


class NERResponse(BaseModel):
    ner_results: List[NERItem] = Field(
        ...,
        description="Named Entity Recognition information",
        example=[
            {"token": "η", "ner_value": "O"},
            {"token": "αργεντινη", "ner_value": "S-ORG"},
            {"token": "κερδισε", "ner_value": "O"},
            {"token": "το", "ner_value": "O"},
            {"token": "παγκοσμιο", "ner_value": "B-EVENT"},
            {"token": "κυπελλο", "ner_value": "E-EVENT"},
            {"token": "το", "ner_value": "O"},
            {"token": "2022", "ner_value": "S-DATE"},
        ],
    )


# @app.post("/ner", response_model=List[NERItem], summary="Named Entity Recognition")
@app.post("/ner", response_model=NERResponse, summary="Named Entity Recognition (NER)")
async def process_ner(
    text: str = Query(
        ...,
        description="The text to process for NER",
        example="Η Αργεντινή κέρδισε το Παγκόσμιο Κύπελλο το 2022",
    ),
):
    """
    The NER endpoint takes Greek text as input and returns a list of dictionaries with the token and the NER value.

    Named Entity Recognition (NER) Labels:
    ```python
        ner_possible_labels = [
            'O', 'S-GPE', 'S-ORG', 'S-CARDINAL', 'B-ORG', 'E-ORG', 'B-DATE', 'E-DATE', 'S-NORP',
            'B-GPE', 'E-GPE', 'S-EVENT', 'S-DATE', 'S-PRODUCT', 'S-LOC', 'I-ORG', 'S-PERSON',
            'S-ORDINAL', 'B-PERSON', 'I-PERSON', 'E-PERSON', 'B-LAW', 'I-LAW', 'E-LAW', 'B-MONEY',
            'I-MONEY', 'E-MONEY', 'B-EVENT', 'I-EVENT', 'E-EVENT', 'B-FAC', 'E-FAC', 'I-DATE',
            'S-PERCENT', 'B-QUANTITY', 'E-QUANTITY', 'B-WORK_OF_ART', 'I-WORK_OF_ART', 'E-WORK_OF_ART',
            'I-FAC', 'S-LAW', 'S-TIME', 'B-LOC', 'E-LOC', 'I-LOC', 'S-FAC', 'B-TIME', 'E-TIME',
            'S-WORK_OF_ART', 'B-PRODUCT', 'E-PRODUCT', 'B-CARDINAL', 'E-CARDINAL', 'S-MONEY',
            'S-LANGUAGE', 'I-TIME', 'I-PRODUCT', 'I-GPE', 'I-QUANTITY', 'B-NORP', 'E-NORP',
            'S-QUANTITY', 'B-PERCENT', 'I-PERCENT', 'E-PERCENT', 'I-CARDINAL', 'B-ORDINAL',
            'I-ORDINAL', 'E-ORDINAL'
        ]
    ```
    """
    try:
        doc = nlp_pos_ner_dp_with_g2g(text)

        # Create a list of dictionaries, each with "token" and "ner_value"
        ner_list = [
            {"token": token.text, "ner_value": token.ner} for token in doc.tokens
        ]

        return {"ner_results": ner_list}

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


# @app.post("/pos", response_model=List[POSItem], summary="Part-of-Speech Tagging")
@app.post("/pos", response_model=POSResponse, summary="Part-of-Speech Tagging (POS)")
async def process_pos(
    text: str = Query(
        ...,
        description="The text to process for POS tagging",
        example="Μου αρέσει να διαβάζω τα post του Andrew Ng στο Twitter.",
    ),
):
    """
    The POS Tagging endpoint analyzes the input text and provides Universal POS (UPOS) tags and detailed morphological features.

    It returns a list of dictionaries with "token", "upos", and "morphological_features" keys.
    The "morphological_features" key contains a dictionary itself with detailed morphological features.

    The UPOS and morphological features are based on the Universal Dependencies (UD) framework: [https://universaldependencies.org/u/pos/](https://universaldependencies.org/u/pos/)

    Complete list of the Universal POS (UPOS) tags and morphological features:
    ```python
        {'ADJ': ['Degree', 'Number', 'Gender', 'Case'],
        'ADP': ['Number', 'Gender', 'Case'],
        'ADV': ['Degree', 'Abbr'],
        'AUX': ['Mood',
                'Aspect',
                'Tense',
                'Number',
                'Person',
                'VerbForm',
                'Voice'],
        'CCONJ': [],
        'DET': ['Number', 'Gender', 'PronType', 'Definite', 'Case'],
        'NOUN': ['Number', 'Gender', 'Abbr', 'Case'],
        'NUM': ['NumType', 'Number', 'Gender', 'Case'],
        'PART': [],
        'PRON': ['Number', 'Gender', 'Person', 'Poss', 'PronType', 'Case'],
        'PROPN': ['Number', 'Gender', 'Case'],
        'PUNCT': [],
        'SCONJ': [],
        'SYM': [],
        'VERB': ['Mood',
                'Aspect',
                'Tense',
                'Number',
                'Gender',
                'Person',
                'VerbForm',
                'Voice',
                'Case'],
        'X': ['Foreign'],
    ```

    ```python
        {'Abbr': ['_', 'Yes'],
        'Aspect': ['Perf', '_', 'Imp'],
        'Case': ['Dat', '_', 'Acc', 'Gen', 'Nom', 'Voc'],
        'Definite': ['Ind', 'Def', '_'],
        'Degree': ['Cmp', 'Sup', '_'],
        'Foreign': ['_', 'Yes'],
        'Gender': ['Fem', 'Masc', '_', 'Neut'],
        'Mood': ['Ind', '_', 'Imp'],
        'NumType': ['Mult', 'Card', '_', 'Ord', 'Sets'],
        'Number': ['Plur', '_', 'Sing'],
        'Person': ['3', '1', '_', '2'],
        'Poss': ['_', 'Yes'],
        'PronType': ['Ind', 'Art', '_', 'Rel', 'Dem', 'Prs', 'Ind,Rel', 'Int'],
        'Tense': ['Pres', 'Past', '_'],
        'VerbForm': ['Part', 'Conv', '_', 'Inf', 'Fin'],
        'Voice': ['Pass', 'Act', '_'],
    ```
    """
    try:
        doc = nlp_pos_ner_dp_with_g2g(text)

        # Create a list of dictionaries, each with "token", "upos", and "morphological_features"
        pos_list = [
            {
                "token": token.text,
                "upos": token.upos,
                "morphological_features": token.feats,
            }
            for token in doc.tokens
        ]

        # return pos_list
        return {"pos_results": pos_list}

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


# @app.post("/dp", response_model=List[DPItem], summary="Dependency Parsing")
@app.post("/dp", response_model=DPResponse, summary="Dependency Parsing (DP)")
async def process_dp(
    text: str = Query(
        ...,
        description="The text to process for Dependency Parsing",
        example="Προτιμώ την πρωινή πτήση από την Αθήνα στη Θεσσαλονίκη",
    ),
):
    """
    The Dependency Parsing endpoint analyzes the syntactic structure of the input text.
    It provides the tokens' (syntactic) heads and dependency relations. A head value of 0 indicates the root.
    More specifically, the endpoint returns a list of dictionaries with "token", "head", and "deprel" keys.

    Dependency Parsing Labels:
    ```python
        dp_possible_labels = ['obl', 'obj', 'dep', 'mark', 'case', 'flat', 'nummod', 'obl:arg', 'punct', 'cop',
        'acl:relcl', 'expl', 'nsubj', 'csubj:pass', 'root', 'advmod', 'nsubj:pass', 'ccomp',
        'conj', 'amod', 'xcomp', 'aux', 'appos', 'csubj', 'fixed', 'nmod', 'iobj', 'parataxis',
        'orphan', 'det', 'advcl', 'vocative', 'compound', 'cc', 'discourse', 'acl', 'obl:agent']
    ```
    """
    try:
        doc = nlp_pos_ner_dp_with_g2g(text)

        # Create a list of dictionaries, each with "token", "head", and "deprel"
        dp_list = [
            {"token": token.text, "head": token.head, "deprel": token.deprel}
            for token in doc.tokens
        ]

        return {"dp_results": dp_list}

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


@app.get("/", include_in_schema=False)
async def root():
    return RedirectResponse(url="/docs#")


if __name__ == "__main__":
    import uvicorn

    uvicorn.run(app)