AutoX_NLP/ nlp_feature.py, OOV问题优化

当前Word2Vec和Glove模型无法处理测试数据中未见过的词，需要对测试数据重新进行词表构建，对整体效果影响较大。
代码链接：[https://github.com/4paradigm/AutoX/blob/master/autox/autox_nlp/feature_engineer/nlp_feature.py](url)

The issue points to nlp_feature.py in autox/autox_nlp/feature_engineer/. Examine how Word2Vec and GloVe models are built and applied, focusing on vocabulary handling for OOV (out-of-vocabulary) words. Look at the test data pipeline to see where the vocabulary is constructed. The goal is to modify the feature engineering to better handle unseen words, likely by adjusting the vocabulary building or embedding fallback strategy.

索引モデルが issue の本文から書いたものです。