Commit 2ec89e50 authored by schaper's avatar schaper
Browse files

Merge branch 'master' of https://gitlab.cl.uni-heidelberg.de/reichelt/semantic-textual-similarity

parents 3d58bf70 69f81891

probing.py

+15 −4
Original line number Diff line number Diff line 
from STSB_Dataset import STSB_Dataset

import pandas as pd

import spacy

import string

import random



df = STSB_Dataset('data/stsbenchmark/sts-test.csv').as_dataframe
@@ -53,8 +55,15 @@ for sent2 in sentences2_nlp: 
    dependency_hierarchy2.append(dep_hierarchy2)





def get_random_string(length):

    letters = string.ascii_lowercase

    result_str = ''.join(random.choice(letters) for i in range(length))

    return result_str



def mask_token(pos_tag):



    df = STSB_Dataset('data/stsbenchmark/sts-test.csv').as_dataframe

    

    for e in range(len(dependency_hierarchy1)):

        mask_token = None

        for layer in dependency_hierarchy1[e]:
@@ -63,7 +72,7 @@ def mask_token(pos_tag): 
            for i in range(len(layer)):

                if layer[i].pos_ == pos_tag:

                    mask_token = layer[i].text

                    df.iloc[e, 0] = df.iloc[e, 0].replace(mask_token, "[MASK]")

                    df.iloc[e, 0] = df.iloc[e, 0].replace(mask_token, get_random_string(random.randrange(1, 9)))

                    break



    for e in range(len(dependency_hierarchy2)):
@@ -74,13 +83,15 @@ def mask_token(pos_tag): 
            for i in range(len(layer)):

                if layer[i].pos_ == pos_tag:

                    mask_token = layer[i].text

                    df.iloc[e, 1] = df.iloc[e, 1].replace(mask_token, "[MASK]")

                    df.iloc[e, 1] = df.iloc[e, 1].replace(mask_token, get_random_string(random.randrange(1, 9)))

                    break

    return df

            



def mask_first(pos_tag):



    df = STSB_Dataset('data/stsbenchmark/sts-test.csv').as_dataframe



    i = -1

    del_rows = []
@@ -91,7 +102,7 @@ def mask_first(pos_tag): 


        if pos_tag in pos_tags_sent:

            e = pos_tags_sent.index(pos_tag)

            df.iloc[i, 0] = df.iloc[i, 0].replace(pos_tag_tokens[e], "[MASK]")

            df.iloc[i, 0] = df.iloc[i, 0].replace(pos_tag_tokens[e], get_random_string(random.randrange(1, 9)))

        else:

            if i not in del_rows:

                del_rows.append(i)
@@ -104,7 +115,7 @@ def mask_first(pos_tag): 


        if pos_tag in pos_tags_sent:

            e = pos_tags_sent.index(pos_tag)

            df.iloc[i, 1] = df.iloc[i, 1].replace(pos_tag_tokens[e], "[MASK]")

            df.iloc[i, 1] = df.iloc[i, 1].replace(pos_tag_tokens[e], random.randrange(1, 9))

        else:

            if i not in del_rows:

                del_rows.append(i)