时间戳

首页 > 解决方案 > 预训练的 Fasttext 模型为词汇表外的单词返回乱码

python - 预训练的 Fasttext 模型为词汇表外的单词返回乱码

问题描述

我在使用预训练的 fasttext.bin 模型（从 https://fasttext.cc/docs/en/crawl-vectors.html检索）时遇到问题。检查 most_similar 是否有 in-vocabulary-words 会返回合理的响应。但是，当检查 most_similar 以查找只有一个字符不同的词汇表外单词时，会返回乱码。

我的问题：这与模型有关，还是我以错误的方式使用它？

from gensim.models.wrappers import FastText
model = FastText.load_fasttext_format('cc.en.300.bin')
model.most_similar("universitet")
[('Universitet', 0.8522759675979614),
 ('högskolan', 0.677900493144989),
 ('Högskola', 0.6725144386291504),
 ('högskola', 0.6724666357040405),
 ('Högskolan', 0.6600401997566223),
 ('Universitetet', 0.6519213318824768),
 ('Høgskolen', 0.647462010383606),
 ('Universiteti', 0.6399329900741577),
 ('forskning', 0.617483377456665),
 ('språk', 0.6172543168067932)]

model.most_similar("universitett")
[('ESTATERETAILCONSUMERPHONESCARSBIKESAPPSINTERNETTABLETSCOMPUTERSSOCIETYPOLITICSLAWCRIMEENVIRONMENTSCIENCEARTSCELEBRITIESSPORTSSPECIALSFIRST',
  0.47905537486076355),
 ('Wikipedia-Page-Suzannah-B-Troy-6-yrs-after-Misogynist-Cyber-Vandalism-Censorship-via-Deletion-on-a-page-about-Censorship-Wikipedia-Agrees-to-retur',
  0.47733378410339355),
 ('DEky4M0BSpUOTPnSpkuL5I0GTSnRI4jMepcaFAoxIoFnX5kmJQk1aYvr2odGBAAIfkECQoABAAsCQAAABAAEgAACGcAARAYSLCgQQEABBokkFAhAQEQHQ4EMKCiQogRCVKsOOAiRocbLQ7EmJEhR4cfEWoUOTFhRIUNE44kGZOjSIQfG9rsyDCnzp0AaMYMyfNjS6JFZWpEKlDiUqALJ0KNatKmU4NDBwYEACH5BAkKAAQALAkAAAAQABIAAAhpAAEQGEiQIICDBAUgLEgAwICHAgkImBhxoMOHAyJOpGgQY8aBGxV2hJgwZMWLFTcCUIjwoEuLBym69PgxJMuDNAUqVDkz50qZLi',
  0.474983274936676),
 ('DEky4M0BSpUOTPnSpkuL5I0GTSnRI4jMepcaFAoxIoFnX5kmJQk1aYvr2odGBAAIfkECQoABAAsCQAAABAAEgAACGcAARAYSLCgQQEABBokkFAhAQEQHQ4EMKCiQogRCVKsOOAiRocbLQ7EmJEhR4cfEWoUOTFhRIUNE44kGZOjSIQfG9rsyDCnzp0AaMYMyfNjS6JFZWpEKlDiUqALJ0KNatKmU4NDBwYEACH5BAUKAAQALAkAAAAQABIAAAhpAAEQGEiQIICDBAUgLEgAwICHAgkImBhxoMOHAyJOpGgQY8aBGxV2hJgwZMWLFTcCUIjwoEuLBym69PgxJMuDNAUqVDkz50qZLi',
  0.47364047169685364),
 ('crescendosexibloguerobateyabsorbersexiindesignabledinerolatifundiosexibrezarcularsutesexirapoplinbrezarcorrentosoVd.lazadareflejoreglafeministabrezarchuzasexiouttiqueblogueroin',
  0.47090965509414673),
 ('QQFZAAEACwAAAAAGQASAAAIjgAJCBQIoGDBgQgTKiwooGHDgwshDgTgsOLDhAAGaAQwUYBBhx85EtS4cWLGjR5JSjxZkgDFkwwLohTJUqTLlANiwvQ4seVNjwwfBoVokKjFo0Jlksz506NFiklZtoQKFSjIoktLVv1YsahSn1WP0vzq02VYoAjJMsVYVKHZrDbdupW6Vq5cunHtRjQoMCAAIfkECRQABAAsCQADAAQABAAACAsABQgkILCgwYEBAQAh',
  0.46747487783432007),
 ('записиТелепрограммаVikerraadioOtseEsilehtJärelkuulamineSaatekavaPodcastidRaadioteaterRaadio',
  0.4659830331802368),
 ('deblogueroreflejoantecedentesexitlacuachebateysuteindesignableabsorbersexilatifundiosexibrezarsutemultiétnicosexiplinrapobrezarcorrentosoVd.lazadafisiochillidomabrezarsico-chuzaoutcolodrablogueroin',
  0.46159273386001587),
 ('2OtseEsilehtJärelkuulamineSaatedPodcastidKlassikaraadioOtseEsilehtJärelkuulamineSaatekavaPodcastidRaadio',
  0.4609595537185669),
 ('leilighetEiendomstypeSelveierleilighetPlass', 0.4550461769104004)]

标签： pythongensimfasttext

解决方案

如果我没记错“gensim.models.wrappers”已被弃用，请尝试使用

从 gensim.models.fasttext 导入 FastText

来源：https ://radimrehurek.com/gensim/models/wrappers/fasttext.html

推荐阅读